Nothing
R/Sim.R
In cascsim: Casualty Actuarial Society Individual Claim Simulator
#' An S4 class to represent a simulation task.
#'
#' @slot startNo The starting simulation index.
#' @slot simNo Number of simulation.
#' @slot lines A string vector to identify the business line(s) to be simulated.
#' @slot types A string vector to identify the claim types to be simulated.
#' @slot iRBNER A Boolean indicating whether IBNER claims need to be simulated.
#' @slot iROPEN A Boolean indicating whether claim reopening needs to be simulated.
#' @slot iIBNR A Boolean indicating whether IBNR claims need to be simulated.
#' @slot iUPR A Boolean indicating whether future claims need to be simulated.
#' @slot claimobjs A list of claim objects.
#' @slot workingFolder A string to specify the working folder where the simulation results will be saved.
#' @slot iCopula A Boolean indicating whether to use copula for frequency simulation.
#' @slot freqCopula Frequency copula.
#' @slot iSummary A Boolean indicating whether to summarzie the simulation results.
#' @slot iReport A Boolean indicating whether to generate an HTML report.
#' @slot iFit A Boolean indicating whether to fit some simulation parameters based on claim data.
#' @slot ncores Number of cores used for simulation.
#' @slot tag A unique tag for the simulation object including date and a random ID.
#' @slot fitfile A string to set the distribution fitting file name. If omitted, a name based on tag will be used.
#' @slot copfile A string to set the copula fitting file name. If omitted, a name based on tag will be used.
#' @slot facfile A string to set the factor fitting file name. Factor table is development year dependant. It could be the probability of zero payment, reopen probability, or loss development factors. If omitted, a name based on tag will be used.
#' @slot fitRpt A string to set the distribution fitting html report file name. If omitted, a name based on tag will be used.
#' @slot simfile A string to set the simulation result file name. If omitted, a name based on tag will be used.
#' @slot sumfile A string to set the summary file name. If omitted, a name based on tag will be used.
#' @slot plog A string to set the parallel run log file name. If omitted, a name based on tag will be used.
#' @slot htmlRpt A string to set the html report name. If omitted, a name based on tag will be used.
#' @slot libpath A string to the R liabrary folder where required packages are installed.
setClass("Simulation",
slots=c(
startNo="numeric",
simNo="numeric",
lines="vector",
types="vector",
# classes="vector",
iRBNER="logical",
iROPEN="logical",
iIBNR="logical",
iUPR="logical",
claimobjs="list",
workingFolder="character",
iCopula="logical",
freqCopula="CopulaObj",
iSummary="logical",
iReport="logical",
iFit="logical",
ncores="numeric",
tag="character",
fitfile="character",
copfile="character",
facfile="character",
fitRpt="character",
simfile="character",
sumfile="character",
plog="character",
htmlRpt="character",
libpath="character"
),
prototype=list(
startNo=1,
simNo=1,
lines=vector(),
types=vector(),
# classes=c("RBNER","ROPEN","IBNR","UPR"),
iRBNER=TRUE,
iROPEN=TRUE,
iIBNR=TRUE,
iUPR=TRUE,
claimobjs=list(),
workingFolder="",
iCopula=FALSE,
freqCopula=new("CopulaObj",param=c(0,0,0),dimension=3),
iSummary=TRUE,
iReport=TRUE,
iFit=TRUE,
ncores=1,
tag=paste0(format(Sys.Date(),"%Y-%m-%d"),"-",round(runif(1,1,10000))),
fitfile="",
copfile="",
facfile="",
fitRpt="",
simfile="",
sumfile="",
plog="",
htmlRpt="",
libpath=""
)
)
#' Claim data fitting analysis at line/type/status level
#' @name claimFitting
#' @param object Simulation object
#' @param claimData claim data including existing claims for RBNER and claim reopenness analysis
#' @param ... Additional parameters that may or may not be used.
#' @rdname claimFitting-methods
#'
#' @importFrom methods new
#'
#' @import grDevices
#'
#' @importFrom R2HTML HTML.title HTML HTMLhr HTMLInsertGraph HTMLCSS
#'
#' @exportMethod claimFitting
setGeneric("claimFitting", function(object, claimData, ...) standardGeneric("claimFitting"))
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param lineList List of business lines to be included in claim fitting;
#' @param typeList List of claim types to be included in claim fitting;
#' @param discreteDist List of discrete distributions to try fitting (report lag, settlemet lag, frequency);
#' @param continuousDist List of continuous distribution to try fitting (severity);
#' @param copulaList List of copula to try fitting;
#' @param fReportLag Boolean variable to indicate whether report lag needs to be fitted;
#' @param fSettlementLag Boolean variable to indicate whether settlement lag needs to be fitted;
#' @param fFrequency Boolean variable to indicate whether monthly frequency needs to be fitted;
#' @param fSeverity Boolean variable to indicate whether severity needs to be fitted;
#' @param fSSRCorrelation Boolean variable to indicate whether copula among severity, report lag and settlement lag needs to be fitted;
#' @param fFreqCorrelation Boolean variable to indicate whether copula among frequencies of business lines needs to be fitted.
#' @param copulaTest Whether to test copula. The testing could take a very long time;
#' @param iTotalLoss Boolean variable to indicate whether total loss before deductible and limit is available for severity fitting;
#' @param fDeductible Boolean variable to indicate whether deductible empirical distribution needs to be fitted;
#' @param fLimit Boolean variable to indicate whether limit empirical distribution needs to be fitted;
#' @param check Boolean variable to indicate whether graph of each tried distribution fitting needs to be generated and saved.
#' @examples
#' library(cascsim)
#' data(claimdata)
#' lines<-c("Auto")
#' types<-c("N")
#' #exposure index
#' index1 <- new("Index",monthlyIndex=c(rep(1,11),cumprod(c(1,rep(1.5^(1/12),11))),
#' cumprod(c(1.5,rep((1.3/1.5)^(1/12),11))),
#' cumprod(c(1.3,rep((1.35/1.3)^(1/12),11))),cumprod(c(1.35,rep((1.4/1.35)^(1/12),11))),rep(1.4,301)))
#' #severity index
#' index2 <- new("Index",monthlyIndex=c(cumprod(c(1,rep(1.03^(1/12),59))),rep(1.03^(5),300)))
#' objan <- new("ClaimType", line="Auto",claimType="N",exposureIndex=index1,severityIndex=index2)
#' objlist <- list(objan)
#' simobj <- new("Simulation",lines=lines,types=types,claimobjs=objlist,iFit=TRUE,
#' iCopula=FALSE, iReport=TRUE, workingFolder=tempdir())
#' simobj <- claimFitting(simobj,claimdata,fSSRCorrelation = FALSE, fSettlementLag = FALSE)
#' @rdname claimFitting-methods
#' @aliases claimFitting,ANY-method
setMethod("claimFitting", signature("Simulation", "data.frame"), function(object, claimData, startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),
lineList = object@lines,
typeList = object@types,
discreteDist = c("Poisson","NegativeBinomial","Geometric"),
continuousDist = c("Normal","Lognormal","Pareto","Weibull","Gamma","Uniform","Exponential"),
copulaList = c("normal"),#c("normal","clayton","gumbel","frank","joe", "t")
fReportLag = TRUE,
fSettlementLag = TRUE,
fFrequency = TRUE,
fSeverity = TRUE,
fSSRCorrelation = TRUE,
fFreqCorrelation = TRUE,
copulaTest = TRUE,
iTotalLoss = TRUE,
fDeductible = TRUE,
fLimit = TRUE,
check = TRUE)
{
tryCatch({
message("Fitting process started.")
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
if (nrow(claimData) > 0) {
claimData[,"occurrenceDate"] = toDate(claimData[,"occurrenceDate"])
claimData[,"reportDate"] = toDate(claimData[,"reportDate"])
claimData[,"settlementDate"] = toDate(claimData[,"settlementDate"])
claimData[,"status"] = toupper(claimData[,"status"])
claimData[,"claimLiability"] = toupper(claimData[,"claimLiability"])
claimData <- claimData[as.Date(claimData[,"occurrenceDate"])>=as.Date(startDate),]
claimData <- claimData[as.Date(claimData[,"reportDate"])<=as.Date(evaluationDate),]
}
if (as.numeric(as.Date(evaluationDate) - as.Date(startDate))<60 & fFrequency == TRUE) {
fFrequency = FALSE
fFreqCorrelation = FALSE
message("Frequency and Frequency Copula Fitting are turned off due to insufficient data.")
message("\n")
}
if(length(object@lines)>0 & length(object@types)>0 & length(object@claimobjs)>0 & nrow(claimData)>0) {
if(object@iFit==TRUE){
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if(object@iReport==TRUE){
if (!dir.exists(paste0(direc,"/fit"))) {dir.create(paste0(direc,"/fit"))}
}
if(object@fitfile==""){
fitname <- paste0("fit",object@tag,".csv")
} else {
fitname <- paste0(object@fitfile,".csv")
}
if(object@fitfile==""){
empname <- paste0("emp",object@tag,".csv")
} else {
empname <- paste0("emp",object@fitfile,".csv")
}
if(object@copfile==""){
copname <- paste0("cop",object@tag,".csv")
} else {
copname <- paste0(object@copfile,".csv")
}
if(object@facfile==""){
facname <- paste0("fac",object@tag,".csv")
} else {
facname <- paste0(object@facfile,".csv")
}
#f <- c("reportLag", "settlementLag", "frequency", "severity", "ssrCorrelation", "freqCorrelation")
empiricals <- data.frame(matrix(seq(0.001,1,0.001),1000,1))
colnames(empiricals) <- "prob"
fitsummary <- data.frame(LoB=character(),
Type=character(),
Fit=character(),
Distribution=character(),
Method=character(),
Parameter=character(),
SD=character(),
p0=numeric(),
DoF=integer(),
ChiSq=double(),
p=double(),
KS=double(),
pks=double(),
loglik=double(),
AIC=double(),
BIC=double(),
stringsAsFactors=FALSE)
rfit<-1
fitsumcop <- data.frame(LoB=character(),
Type=character(),
Fit=character(),
Copula=character(),
Method=character(),
Parameter=character(),
SD=character(),
DoF=integer(),
Sn=double(),
p=double(),
stringsAsFactors=FALSE)
rcop<-1
fitsumfac <- data.frame(LoB=character(),
Type=character(),
Fit=character(),
Year=character(),
MeanList=double(),
VolList=double(),
stringsAsFactors=FALSE)
rfac<-1
for(l in lineList){#object@lines
for(t in object@types){#object@lines
fitdata <- claimData[claimData[,"LoB"]==l,]
fitdata <- fitdata[fitdata[,"Type"]==t,]
obji<-1
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
obji <- obji + 1
}
if(nrow(fitdata)>5 & co@line==l & co@claimType==t){
if(fReportLag==TRUE){
f <- "reportLag"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
reportlags <- as.numeric(as.Date(fitdata[,"reportDate"])-as.Date(fitdata[,"occurrenceDate"]))
reportlags <- ifelse(reportlags==0, runif(length(reportlags)),reportlags)
if(TRUE){#(objName(co@reportLag)!="Empirical"){
xFit <- new("FitDist", observation=as.data.frame(reportlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
bestBIC <- 1e10
irlso <- 0
for (i in c(1:length(continuousDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(continuousDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
irlso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,continuousDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",continuousDist[i]," distribution"))
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
#message(toString(xFit@fitted))
}
if (irlso == 1) {
co@reportLag <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@reportLag)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@reportLag@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=as.data.frame(reportlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@reportLag <- xFit@fitted
#xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=as.data.frame(reportlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
message("\n")
}
if(fFrequency==TRUE){
f <- "frequency"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
rawdata <- as.data.frame(as.Date(fitdata$occurrenceDate))
colnames(rawdata)<-"occurrenceDate"
if(TRUE){#(objName(co@frequency)!="Empirical"){
xFit <- new("FitDist", observation=rawdata, trend=co@exposureIndex,startDate = startDate, endDate= evaluationDate, method="mle",ifreq=TRUE,idate=TRUE, freq="Monthly", iLag=TRUE, reportLag=co@reportLag)
xFit <- setFitdata(xFit)
bestBIC <- 1e10
ifso <- 0
for (i in c(1:length(discreteDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(discreteDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",discreteDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",discreteDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
ifso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,discreteDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,discreteDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,discreteDist[i],".jpg"))
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",discreteDist[i]," distribution"))
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
#message(toString(xFit@fitted))
}
if (ifso == 1) {
co@frequency <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@frequency)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@frequency@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=rawdata, trend=co@exposureIndex,startDate = startDate, method="mle",ifreq=TRUE,idate=TRUE, freq="Monthly", iLag=TRUE, reportLag=co@reportLag)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@frequency <- xFit@fitted
#xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=rawdata, trend=co@exposureIndex,startDate = startDate, endDate= evaluationDate, method="mle",ifreq=TRUE,idate=TRUE, freq="Monthly", iLag=TRUE, reportLag=co@reportLag)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
message("\n")
}
if(fSettlementLag==TRUE){
fitdatacls <- fitdata[fitdata$status=="CLOSED",]
f <- "settlementLag"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
settlementlags <- as.numeric(as.Date(fitdatacls[,"settlementDate"])-as.Date(fitdatacls[,"reportDate"]))
rm(fitdatacls)
settlementlags <- settlementlags[!is.na(settlementlags)]
settlementlags <- ifelse(settlementlags==0, runif(length(settlementlags)),settlementlags)
if(TRUE){#(objName(co@settlementLag)!="Empirical"){
xFit <- new("FitDist", observation=as.data.frame(settlementlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
bestBIC <- 1e10
islso <- 0
for (i in c(1:length(continuousDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(continuousDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
islso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,continuousDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",continuousDist[i]," distribution"))
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
}
if (islso == 1) {
co@settlementLag <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@settlementLag)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@settlementLag@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=as.data.frame(settlementlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@settlementLag <- xFit@fitted
#xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=as.data.frame(settlementlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
message("\n")
}
if(fDeductible==TRUE){
f <- "Deductible"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
deductibles <- as.numeric(fitdata[,"Deductible"])
deductibles <- ifelse(is.na(deductibles),0,deductibles)
deductibles <- ifelse(deductibles<0,0,deductibles)
xFit <- new("FitDist", observation=as.data.frame(deductibles), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
co@deductible <- xFit@fitted
rfit <- rfit + 1
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=240,height=240)
CDFPlot(xFit)
dev.off()
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical",".jpg"))
par(mfrow = c(1, 1))
CDFPlot(xFit)
dev.off()
}
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
message("\n")
}
if(fLimit==TRUE){
f <- "Limit"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
limits <- as.numeric(fitdata[,"Limit"])
limits <- ifelse(is.na(limits),1e10,limits)
limits <- ifelse(limits<0,1e10,limits)
xFit <- new("FitDist", observation=as.data.frame(limits), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
co@limit <- xFit@fitted
rfit <- rfit + 1
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=240,height=240)
CDFPlot(xFit)
dev.off()
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical",".jpg"))
par(mfrow = c(1, 1))
CDFPlot(xFit)
dev.off()
}
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
message("\n")
}
if(fSeverity==TRUE){
f <- "severity"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
nclosed <- nrow(fitdata[fitdata[,"status"]=="CLOSED",])
nvalid <- nrow(fitdata[fitdata[,"claimLiability"]==TRUE & fitdata[,"status"]=="CLOSED",])
aggP0 <- 1-nvalid/nclosed
p0Set <- fitdata[fitdata[,"status"]=="CLOSED",]
p0Set[,"settlementYears"] <- ceiling(as.numeric(as.Date(p0Set[,"settlementDate"]) - as.Date(p0Set[,"occurrenceDate"]))/365)
p0Set <- p0Set[,colnames(p0Set) %in% c("settlementYears","claimLiability")]
#p0Set <- read.csv("C:/temp/CAS/pset.csv")
yclosed <- aggregate(claimLiability ~ settlementYears, data=p0Set, FUN="length")
if(nrow(p0Set[p0Set$claimLiability==FALSE,])>0){
yinvalid <- aggregate(claimLiability ~ settlementYears, data=p0Set[p0Set$claimLiability==FALSE,], FUN="length")
vp0 <- rep(0,max(yclosed$settlementYears))
for (i in c(1:length(vp0))) {
denominator <- yclosed[yclosed$settlementYears == i,]$claimLiability[1]
numerator <- yinvalid[yinvalid$settlementYears == i,]$claimLiability[1]
vp0[i] <- numerator/denominator
vp0[i] <- min(1,vp0[i])
}
vp0[is.na(vp0)] <- 0
} else {
vp0 <- rep(0,max(yclosed$settlementYears))
}
for (i in c(1:length(vp0))) {
fitsumfac[rfac,] <- c(l,t,"p0",i,vp0[i],0)
rfac <- rfac + 1
}
co@p0 <- new("DevFac",FacID=paste0(l,t,"p0"),FacModel= FALSE,meanList =vp0,volList =rep(0,length(vp0)))
if(co@sdata == "CLOSED") {
fitdata <- fitdata[fitdata[,"claimLiability"]==TRUE & fitdata[,"status"]=="CLOSED",]
} else {
fitdata <- fitdata[fitdata[,"claimLiability"]==TRUE,]
}
if (iTotalLoss == TRUE) {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","totalLoss")][,c("settlementDate","totalLoss")]
} else {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","incurredLoss","Deductible","Limit")][,c("settlementDate","incurredLoss","Deductible","Limit")]
}
if (iTotalLoss == TRUE) {
incurredlosses[,2] <- pmax(0.01,incurredlosses[,2])
}
if(TRUE){#objName(co@reportLag)!="Empirical"){
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
if (iTotalLoss == FALSE) {
xFit@iDL = TRUE
}
xFit <- setFitdata(xFit)
losses <- xFit@fitdata
bestBIC <- 1e10
isso <- 0
for (i in c(1:length(continuousDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(continuousDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
isso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- aggP0
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,continuousDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
par(mfrow = c(1, 1))
if (iTotalLoss == FALSE) {
plotText(paste0("Data cannot fit to truncated ",continuousDist[i]," distribution"))
} else {
plotText(paste0("Data cannot fit to ",continuousDist[i]," distribution"))
}
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
}
if (isso == 1) {
co@severity <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@severity)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@severity@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@severity <- xFit@fitted
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
fitdata <- fitdata[fitdata[,"claimLiability"]==TRUE & fitdata[,"status"]=="CLOSED",]
reportlags <- as.numeric(as.Date(fitdata[,"reportDate"])-as.Date(fitdata[,"occurrenceDate"]))
reportlags <- ifelse(reportlags==0, runif(length(reportlags)),reportlags)
settlementlags <- as.numeric(as.Date(fitdata[,"settlementDate"])-as.Date(fitdata[,"reportDate"]))
settlementlags <- ifelse(settlementlags==0, runif(length(settlementlags)),settlementlags)
if (iTotalLoss == TRUE) {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","totalLoss")][,c("settlementDate","totalLoss")]
} else {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","incurredLoss","Deductible","Limit")][,c("settlementDate","incurredLoss","Deductible","Limit")]
}
if (iTotalLoss == TRUE) {
incurredlosses[,2] <- pmax(0.01,incurredlosses[,2])
}
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
if (iTotalLoss == FALSE) {
xFit@iDL = TRUE
}
xFit <- setFitdata(xFit)
losses <- xFit@fitdata
message("\n")
}
if(co@iCopula==TRUE & fReportLag==TRUE & fSettlementLag==TRUE & fSeverity==TRUE & fSSRCorrelation==TRUE){
f <- "ssrCorrelation"
message(paste0("Start Fitting Line:",l," Type:",t," ","Correlation among severity, report lag and settlement lag"))
x <- cbind(losses,settlementlags,reportlags)
if (nrow(x) > 2000) {
x <- x[sample(nrow(x), 2000), ]
}
if(nrow(x)>10) {
dist1 <- co@severity
dist2 <- co@settlementLag
dist3 <- co@reportLag
bestP <- 0
issrso <- 0
for (i in c(1:length(copulaList))){
if(copulaList[i]=="normal"){
nom.cop <- new("CopulaObj", type="normal",param=c(0.5,0.5,0.5),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,observation=x,fittest=copulaTest)
} else if(copulaList[i]=="t"){
nom.cop <- new("CopulaObj", type="t",param=c(0.5,0.5,0.5),df=5, marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,observation=x,fittest=copulaTest)
} else {
nom.cop <- new("CopulaObj", type=copulaList[i],param=c(3),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,observation=x,fittest=copulaTest)
}
so <- tryCatch({
if(nom.cop@type=="t") {nom.cop@fittest=TRUE} else {nom.cop@fittest=TRUE}#FALSE
nom.cop <- copulaFitErr(nom.cop)
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula failed to fit"))
-1
}
)
if(so==1) {
issrso <- 1
fitsumcop[rcop,] <- c(l,t,f,nom.cop@coutput)
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=240)
copulaFitPlot(nom.cop)
dev.off()
}
} else {
fitsumcop[rcop,] <- c(l,t,f,copulaList[i],"mpl",rep(NA,5))
nom.cop@coutput <- data.frame(matrix(c(copulaList[i],"mpl",rep(NA,5)),1,7))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=480)
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",copulaList[i]," copula"))
dev.off()
}
}
rcop <- rcop+1
if (i==1) {
bestfit <- nom.cop
if (!is.na(nom.cop@coutput[1,7])) {bestP <- as.numeric(nom.cop@coutput[1,7])}
} else if ((!is.na(nom.cop@coutput[1,7])) & (as.numeric(nom.cop@coutput[1,7])>bestP)){
bestfit <- nom.cop
bestP <- as.numeric(nom.cop@coutput[1,7])
}
}
if (issrso == 1) {
bestfit@fitsucc = TRUE
co@ssrCopula <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@ssrCopula)))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
copulaFitPlot(bestfit)
dev.off()
} else {
nom.cop <- new("CopulaObj", type="normal",param=c(0,0,0),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,fittest=copulaTest)
nom.cop@coutput = data.frame(Copula=character(),
Method=character(),
Parameter=character(),
SD=character(),
DoF=integer(),
Sn=double(),
p=double(),
stringsAsFactors=FALSE)
nom.cop@coutput[1,] = c("normal", "mpl", NA, NA, 3, NA, NA)
co@ssrCopula <- nom.cop
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No copula is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data cannot fit to any tested copula"))
dev.off()
}
message("\n")
} else {
dist1 <- co@severity
dist2 <- co@settlementLag
dist3 <- co@reportLag
nom.cop <- new("CopulaObj", type="normal",param=c(0,0,0),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,fittest=copulaTest)
nom.cop@coutput = data.frame(Copula=character(),
Method=character(),
Parameter=character(),
SD=character(),
DoF=integer(),
Sn=double(),
p=double(),
stringsAsFactors=FALSE)
nom.cop@coutput[1,] = c("normal", "mpl", NA, NA, 3, NA, NA)
co@ssrCopula <- nom.cop
message(paste0("Severity, settlement lag and report lag copula fitting is turned off due to insufficient data for line " , l, " type " , t))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data are insufficient fit to any tested copula"))
dev.off()
message("\n")
}
}
if(co@line==l & co@claimType==t) {object@claimobjs[[obji]]<-co}
} else {
message (paste0("Nothing to fit for line " , l, " type " , t))
message("\n")
}
}
}
if(fFreqCorrelation==TRUE & object@iCopula==TRUE){
tmp <- cbind(as.character(claimData$LoB),as.character(claimData$Type),claimData$occurrenceDate,as.character(substr(as.character(claimData$occurrenceDate),1,7)))
colnames(tmp) <- c("line","type","od","ym")
tmp <- aggregate(od ~ ym+line+type, data=tmp, FUN="length")
tmp <- cbind(tmp,as.numeric(substr(tmp$ym,1,4)),as.numeric(substr(tmp$ym,6,7)))
colnames(tmp) <- c("ym","line","type","od","year","mth")
for(l in object@lines){
for(t in object@types){
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if(co@line==l & co@claimType==t){
startYear <- as.numeric(substr(as.character(co@exposureIndex@startDate),1,4))
startMonth <- as.numeric(substr(as.character(co@exposureIndex@startDate),6,7))
endYear <- as.numeric(substr(as.character(evaluationDate),1,4))
endMonth <- as.numeric(substr(as.character(evaluationDate),6,7))
tmp$od <- ifelse((tmp$line==l & tmp$type==t),tmp$od/Probability(co@reportLag,(endYear-tmp$year)*365+(endMonth-tmp$mth)*30+15),tmp$od)
di <- (tmp$year - startYear)*12+(tmp$mth - startMonth)+1
di <- ifelse(di>360,360,ifelse(di<1,1,di))
tmp$od <- ifelse((tmp$line==l & tmp$type==t),tmp$od/co@exposureIndex@monthlyIndex[di],tmp$od)
tmp$od <- round(tmp$od)
}
}
}
tmp <- aggregate(od ~ ym+line, data=tmp, FUN="sum")
uym<-as.data.frame(as.character(unique(tmp$ym)))
colnames(uym)<-"uym"
for(l in object@lines){
uym[,l] <- 0
}
for(j in c(1:nrow(uym))){
for(l in object@lines){
uym[j,l] <- if(nrow(tmp[tmp$ym==uym$uym[j] & tmp$line==l,])==0) {0} else {tmp[tmp$ym==uym$uym[j] & tmp$line==l,]$od}
}
}
uym <- na.omit(uym[uym[2:ncol(uym)]>0,])
f <- "freqCorrelation"
l <- "Total"
t <- "Total"
message(paste0("Start Fitting frequency correlation among business lines"))
x <- data.matrix(uym[,-1])
if (nrow(x)>=10) {
bestP <- 0
ifcso <- 0
for (i in c(1:length(copulaList))){
if(copulaList[i]=="normal"){
nom.cop <- new("CopulaObj", type="normal",param=rep(0.5,(length(object@lines)-1)*length(object@lines)/2),dimension=length(object@lines),observation=x,fittest=copulaTest)
} else if(copulaList[i]=="t"){
nom.cop <- new("CopulaObj", type="t",param=rep(0.5,(length(object@lines)-1)*length(object@lines)/2),df=5, dimension=length(object@lines),observation=x,fittest=copulaTest)
} else {
nom.cop <- new("CopulaObj", type=copulaList[i],param=c(3),dimension=length(object@lines),observation=x,fittest=copulaTest)
}
so <- tryCatch({
nom.cop <- copulaFitErr(nom.cop)
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula failed to fit"))
-1
}
)
if(so==1) {
ifcso <- 1
fitsumcop[rcop,] <- c(l,t,f,nom.cop@coutput)
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=240)
copulaFitPlot(nom.cop)
dev.off()
}
} else {
fitsumcop[rcop,] <- c(l,t,f,copulaList[i],"mpl",rep(NA,5))
nom.cop@coutput <- data.frame(matrix(c(copulaList[i],"mpl",rep(NA,5)),1,7))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=480)
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",copulaList[i]," copula"))
dev.off()
}
}
rcop <- rcop+1
if (i==1) {
bestfit <- nom.cop
if (!is.na(nom.cop@coutput[1,7])) {bestP <- as.numeric(nom.cop@coutput[1,7])}
} else if ((!is.na(nom.cop@coutput[1,7])) & (as.numeric(nom.cop@coutput[1,7])>bestP)){
bestfit <- nom.cop
bestP <- as.numeric(nom.cop@coutput[1,7])
}
}
if (ifcso == 1) {
object@freqCopula <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(object@freqCopula)))
png(filename = paste0(direc,"/fit/","fit",object@tag,f,".png"),width=480,height=240)
copulaFitPlot(bestfit)
dev.off()
} else {
nom.cop <- new("CopulaObj", type="normal",param=rep(0,(length(object@lines)-1)*length(object@lines)/2),dimension=length(object@lines),fittest=copulaTest)
nom.cop@coutput <- data.frame(matrix(c("normal","mpl",rep(NA,5)),1,7))
object@freqCopula <- nom.cop
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No copula is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data cannot fit to any tested copula"))
dev.off()
}
} else {
nom.cop <- new("CopulaObj", type="normal",param=rep(0,(length(object@lines)-1)*length(object@lines)/2),dimension=length(object@lines),fittest=copulaTest)
nom.cop@coutput <- data.frame(matrix(c("normal","mpl",rep(NA,5)),1,7))
object@freqCopula <- nom.cop
message(paste0("Frequency copula fitting is turned off due to insufficient data"))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data is insufficient to fit to any tested copula"))
dev.off()
fFreqCorrelation = FALSE
}
}
if (check == TRUE) {
jpeg(filename = paste0(direc,"/fit/",object@fitfile,"na.jpg"),width=480,height=480)
par(mfrow = c(1, 1))
plotText(paste0("Not Fitted"))
dev.off()
jpeg(filename = paste0(direc,"/fit/",object@fitfile,"halfna.jpg"),width=480,height=240)
par(mfrow = c(1, 1))
plotText(paste0("Not Fitted"))
dev.off()
}
if(nrow(fitsummary)>=1) {write.table(fitsummary,paste0(direc,"/",fitname),row.names=FALSE, sep=",")}
if(ncol(fitsumfac)>=1) {write.table(fitsumfac,paste0(direc,"/",facname),row.names=FALSE, sep=",")}
if(nrow(fitsumcop)>=1) {write.table(fitsumcop,paste0(direc,"/",copname),row.names=FALSE, sep=",")}
if(ncol(empiricals)>=1) {write.table(empiricals,paste0(direc,"/",empname),row.names=FALSE, sep=",")}
if(object@iReport==TRUE){
#require(R2HTML)
pckgdir <- find.package("cascsim")
pckgdir <- paste0(pckgdir,"/css/")
if (!dir.exists(paste0(direc,"/fit"))) {dir.create(paste0(direc,"/fit"))}
copied <- file.copy(file.path(pckgdir, "report.css"), file.path(paste0(direc,"/fit/"),"report.css"))
if(object@fitRpt==""){
fitrpt <- paste0("sim",object@tag)
} else {
fitrpt <- paste0(object@fitRpt)
}
if (file.exists(paste0(direc,"/fit/",fitrpt,".html"))) {file.remove(paste0(direc,"/fit/",fitrpt,".html"))}
ReportBegin <- function(file, title, directory = direc) {
file = file.path(paste0(directory,"/fit"), file)
cat(paste("<html><head><title>", title, "</title></head>", "<body bgcolor=#D0D0D0>", sep = ""), file = file, append = FALSE)
}
ReportEnd <- function(file, directory = direc) {
file = file.path(paste0(directory,"/fit"), file)
cat("<hr size=1></body></html>", file = file, append = TRUE)
}
RptFetch<-function(line,type,dist){
if(dist=="ssrCorrelation" | dist=="freqCorrelation"){
result <- fitsumcop[fitsumcop$LoB==line & fitsumcop$Type==type & fitsumcop$Fit==dist,]
}else{
result <- fitsummary[fitsummary$LoB==line & fitsummary$Type==type & fitsummary$Fit==dist,]
}
rownames(result)<-NULL
return(result)
}
vecFlatten<-function(vec){
result<-""
if(length(vec)>1){
for(i in c(1:(length(vec)-1))){
result<-paste0(result,vec[i],", ")
}
result <- paste0(result,"and ",vec[length(vec)])
} else {
result <- vec[1]
}
return(result)
}
getCopulaInfo<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-object@lines
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
getCopulaSR<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-c("Severity","ReportLag","SettlementLag")
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
findClaimObj <- function(line,type,objpool){
for(co in objpool){
if(co@line==line & co@claimType==type) {
result <- co
break
} else {
result <- NULL
}
}
return(result)
}
HTMLcontent <- function(file, append = TRUE, directory = direc) {
file = file.path(paste0(directory,"/fit"), file)
cat("\n", file = file, append = append)
HTML.title("Claim Data Fitting Report", file=file, HR=1)
HTML(paste0("Claim Data Start Date: ",startDate,"; Claim Data Ending Date: ",evaluationDate), file=file)
HTML(paste0("Simulation Task ID: ",object@tag), file=file)
HTML(paste0("Report generated at ",date()), file=file)
HTMLhr(file = file)
HTML.title("Links",file = file, HR=4)
HTML(paste0("<a href=\"#tp\">Total Portfolio</a>"),file=file)
for(l in object@lines){
#HTML(paste0("<a href=\"#",l,"\">",l,"-Total","</a>"),file=file)
for(t in object@types){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
s<-paste0("<a href=\"#",l,t,"\">",l,"-",t,"</a>")
HTML(s,file=file)
}
}
}
HTMLhr(file = file)
HTML(paste0("<a name=\"tp\"></a>"), file = file)
HTML.title("Total Portfolio",file = file, HR=2)
HTML(paste0("The portfolio includes ",length(object@lines)," business line(s): ",vecFlatten(object@lines),"."),file=file)#,". Each business line may contain ",length(object@types)," type(s): ",vecFlatten(object@types)
if(fFreqCorrelation==TRUE & object@iCopula==TRUE){
HTML.title("Frequency Copula Among Business Lines",file = file, HR=3)
if (ifcso == 1) {
HTML(paste0("Best Copula: ",getCopulaInfo(object@freqCopula)$out), file = file)
if (!is.null(getCopulaInfo(object@freqCopula)$outm)){
HTML(getCopulaInfo(object@freqCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,"freqCorrelation.png"),Caption = "Copula Fitting Plot",file = file)
} else {
HTML(paste0("No tested copula is found appropriate."),file=file)
}
HTML("Tested Copulas", file = file)
HTML(RptFetch("Total","Total","freqCorrelation")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLhr(file = file)
for(l in object@lines){
HTML(paste0("<a name=\"",l,"\"></a>"), file = file)
for(t in object@types){
if (!is.null(findClaimObj(l,t,object@claimobjs))){
if(fFrequency==TRUE){
HTML(paste0("<a name=\"",l,t,"\"></a>"), file = file)
HTML.title(paste0(l,"-",t),file = file, HR=2)
HTML.title(paste0("Frequency"),file = file, HR=3)
#if (ifso == 1) {
if (findClaimObj(l,t,object@claimobjs)@frequency@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@frequency)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"frequency.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"frequency")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
}
if(fSeverity==TRUE){
HTML.title(paste0("Severity"),file = file, HR=3)
#if (isso == 1) {
if (findClaimObj(l,t,object@claimobjs)@severity@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"severity.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"severity")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Probability of Zero Payment by Development Year(p0): ",findClaimObj(l,t,object@claimobjs)@p0@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@p0), file = file, innerBorder=0.5, row.names=TRUE)
HTMLhr(file = file)
}
}
if(fReportLag==TRUE){
HTML.title(paste0("Report Lag"),file = file, HR=3)
#if (irlso == 1) {
if (findClaimObj(l,t,object@claimobjs)@reportLag@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reportLag)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"reportLag.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"reportLag")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
}
if(fSettlementLag==TRUE){
HTML.title(paste0("Settlement Lag"),file = file, HR=3)
#if (islso == 1) {
if (findClaimObj(l,t,object@claimobjs)@settlementLag@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@settlementLag)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"settlementLag.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"settlementLag")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
}
if(fDeductible==TRUE){
HTML.title(paste0("Deductible Empirical Distribution"),file = file, HR=3)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"Deductible.png"),Caption = "",file = file,WidthHTML=250)
HTMLhr(file = file)
}
if(fLimit==TRUE){
HTML.title(paste0("Limit Empirical Distribution"),file = file, HR=3)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"Limit.png"),Caption = "",file = file,WidthHTML=250)
HTMLhr(file = file)
}
if(findClaimObj(l,t,object@claimobjs)@iCopula==TRUE & fReportLag==TRUE & fSettlementLag==TRUE & fSeverity==TRUE & fSSRCorrelation==TRUE){
HTML.title(paste0("Severity, Report Lag and Settlement Lag Copula"),file = file, HR=3)
#if (issrso == 1) {
if (findClaimObj(l,t,object@claimobjs)@ssrCopula@fitsucc == TRUE) {
HTML(paste0("Best Copula: ",getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$out), file = file)
HTML(getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"ssrCorrelation.png"),Caption = "",file = file)
HTML("Tested Copulas", file = file)
HTML(RptFetch(l,t,"ssrCorrelation")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested copula is found appropriate."),file=file)
}
HTMLhr(file = file)
}
}
}
}
message(paste("Report generated: ",file," at ",date(), "\n", sep = ""))
}
Report <- function(file) {
ReportBegin(file,"Claim Data Fitting Report")
if(copied == TRUE | file.exists(paste0(direc,"/fit/","report.css"))) {
HTMLCSS(file = paste0(direc,"/fit/",file), CSSfile = paste0(direc,"/fit/","report.css"))
}
HTMLcontent(file)
ReportEnd(file)
}
big_mark_opt <- getOption("R2HTML.format.big.mark")
decimal_mark_opt <- getOption("R2HTML.format.decimal.mark")
options("R2HTML.format.big.mark"=",")
options("R2HTML.format.decimal.mark"=".")
Report(paste0(fitrpt,".html"))
on.exit(options("R2HTML.format.big.mark"=big_mark_opt))
on.exit(options("R2HTML.format.decimal.mark"=decimal_mark_opt), add = TRUE)
}
} else {
stop("Reset iFit of the simulation object to TRUE to fit frequency, severity, report lag, settlement lag and frequency correlation to claim data.")
}
message("Claim data fitting done")
return(object)
} else {
message("Nothing to fit or no claim data.")
message("Claim data fitting done")
}
}, error = function(err){
message("Something is wrong. Please check the error messages")
if (exists("l") & exists("t") & exists("f")){
message(paste0(">>>Critical Error for claim distribution fitting: ",err," Line-", l, " Type-", t, " Fitting-", f))
} else {
message(paste0(">>>Critical Error for claim distribution fitting: ",err))
}
message("Claim data fitting done")
gc()
return(-1)
})
})
#' Claim simulation at line/type/status level
#' @name claimSimulation
#' @param object Simulation object
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' library(cascsim)
#' data(claimdata)
#' lines <- c("Auto")
#' types <- c("N")
#' AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' AutoN@frequency <- new("Poisson", p1 =50)
#' AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' AutoN@reportLag <- new("Exponential", p1 =0.1)
#' AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' AutoN@iCopula <- TRUE #use copula
#' AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' param = c(0.1,0.2,0.1))#A Gaussian Copula
#' AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' paras =c(5,1.5,0.005,1.2,3))
#' AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' simobj <- new("Simulation", lines=lines, types=types,
#' claimobjs= list(AutoN),workingFolder=tempdir())
#' simobj@simNo <- 1
#' simobj@iRBNER <-FALSE
#' simobj@iROPEN <-FALSE
#' simobj@iIBNR <-TRUE
#' simobj@iUPR <-FALSE
#' simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' @rdname claimSimulation-methods
#'
#' @importFrom utils write.table
#' @importFrom parallel clusterSetRNGStream makeCluster clusterExport parLapply stopCluster
#'
#' @exportMethod claimSimulation
setGeneric("claimSimulation", function(object, ...) standardGeneric("claimSimulation"))
#' @param claimData claim data including existing claims for RBNER and claim reopenness analysis;
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims).
#' @param append Boolean variable to indicate whether existing simulation results need to be kept.
#' @rdname claimSimulation-methods
#' @aliases claimSimulation,ANY-method
setMethod("claimSimulation", signature("Simulation"), function(object, claimData=data.frame(), startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"),append=TRUE)
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
#rn<-round(runif(1,1,10000))
if(object@simfile==""){
filename <- paste0("sim",object@tag,".csv")
} else {
filename <- paste0(object@simfile,".csv")
}
if(object@plog==""){
mcfile <- paste0("mcsim",object@tag,".txt")
} else {
mcfile <- paste0(object@plog,".txt")
}
simdata = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
iclass <- object@iRBNER | object@iROPEN | object@iIBNR | object@iUPR
if(length(object@lines)>0 & length(object@types)>0 & length(object@claimobjs)>0 & iclass) {
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if (file.exists(mcfile)) {file.remove(mcfile)}
if (append==FALSE & file.exists(filename)) {file.remove(filename)}
if(evaluationDate < startDate | futureDate < startDate){
stop("Evaluation date cannot be earlier than the start date.")
} else {
nmonths <- (as.numeric(substr(as.character(futureDate),1,4))-as.numeric(substr(as.character(startDate),1,4)))*12+as.numeric(substr(as.character(futureDate),6,7))-as.numeric(substr(as.character(startDate),6,7))+1
}
monthlyfreq <- function(mthfreq,startDate,endDate,objRptLag,iRptLag=FALSE) {
startyear<-as.numeric(substr(as.character(startDate),1,4))
startmonth<-as.numeric(substr(as.character(startDate),6,7))
endyear<-as.numeric(substr(as.character(endDate),1,4))
endmonth<-as.numeric(substr(as.character(endDate),6,7))
nmth<-(endyear-startyear)*12+endmonth-startmonth+1
istart <- max(1,(startyear-as.numeric(substr(as.character(startDate),1,4)))*12+startmonth-as.numeric(substr(as.character(startDate),6,7))+1)
iend <- min(istart + nmth - 1,nmonths)
result <- mthfreq[istart:iend]
if (iRptLag == FALSE) {
return(round(result))
} else {
for (i in c(1:(iend-istart+1))){
prob <- mean(Probability(objRptLag,(nmth-istart-i+1)*30+c(1:30)))
result[i] <- result[i]*(1-prob)
}
return(round(result))
}
}
if(object@ncores>1 & object@simNo >1){
#require(parallel)
pC <- makeCluster(object@ncores, outfile = mcfile)
vsim <- rep(0,object@ncores)
vsims <- rep(0,object@ncores)
vobj<-list()
if (object@ncores > object@simNo) {
object@ncores <- max(object@simNo,2)
}
for (j in c(1:object@ncores)){
vobj<- c(vobj,object)
vsim[j] <- floor(object@simNo/object@ncores)*j
vsims[j] <- floor(object@simNo/object@ncores)*(j-1)+1
}
vsim[object@ncores] <- object@simNo
#set.seed(123)
clusterSetRNGStream (cl = pC)
clusterExport(cl=pC, varlist=c("vsim", "vsims", "vobj", "claimData", "nmonths", "startDate", "evaluationDate", "futureDate", "monthlyfreq", "filename"), envir=environment())
msim <- function(i){
if(nchar(object@libpath)>0 & dir.exists(object@libpath)){
.libPaths(object@libpath)
library(cascsim,lib.loc=object@libpath)
}
simdata = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
if(object@simfile==""){
filenamei <- paste0("sim",object@tag,"c",i,".csv")
} else {
filenamei <- paste0(object@simfile,"c",i,".csv")
}
for(isim in c(vsims[i]:vsim[i])){
object<-vobj[[i]]
message(paste0("Worker ",i,",Simulation ",isim+object@startNo-1," started at ",date()))
if(object@iCopula==TRUE) {
freqs <- copulaSample(object@freqCopula,nmonths)
}
for(l in object@lines){
for(t in object@types){
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if(co@line==l & co@claimType==t){
co@simno<-as.character(isim+object@startNo-1)
if(object@iIBNR==TRUE){
if(object@iCopula==TRUE){
co@IBNRfreqIndex@startDate <- startDate
lid <- match(l,object@lines)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),startDate,evaluationDate,co@reportLag,TRUE))
} else {
co@IBNRfreqIndex@startDate <- startDate
mthfreq <- doSample(co@frequency,nmonths)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(mthfreq,startDate,evaluationDate,co@reportLag,TRUE))
}
}
if(object@iUPR==TRUE){
if(object@iCopula==TRUE){
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
lid <- match(l,object@lines)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
} else {
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
mthfreq <- doSample(co@frequency,nmonths)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(mthfreq,as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
}
}
stmp = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
co@iRBNER = object@iRBNER
co@iROPEN = object@iROPEN
co@iIBNR = object@iIBNR
co@iUPR = object@iUPR
stmp<-rbind(stmp,claimSample(co,claimData,startDate,evaluationDate)) #futureDate
if(file.exists(paste0(direc,"/",filenamei))){
write.table(stmp,paste0(direc,"/",filenamei),row.names=FALSE, sep=",",append = TRUE, col.names = FALSE)
}else{
write.table(stmp,paste0(direc,"/",filenamei),row.names=FALSE, sep=",",append = TRUE, col.names = TRUE)
}
simdata <- rbind(simdata,stmp)
}
}
}
# if (isim %% 10 == 0) {
message(paste0("Worker",i,",Simulation ",isim+object@startNo-1," ended at ",date()))
message(">>>Simulation is finished successfully<<<")
}
#file.rename("simdatatmp.csv", filename)
return(simdata)
}
#ptm<-proc.time()
#clusterExport(pC, "ptm")
msimdata <- parLapply(pC, 1:(vobj[[1]]@ncores), msim)
simdata = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
Paid=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
for (k in c(1:length(msimdata))){
simdata<-rbind(simdata,msimdata[[k]])
}
stopCluster (pC)
rm(msimdata)
gc()
if(file.exists(paste0(direc,"/",filename))){
write.table(simdata,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = FALSE)
}else{
write.table(simdata,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = TRUE)
}
for(i in c(1:object@ncores)){
if(object@simfile==""){
if(file.exists(paste0("sim",object@tag,"c",i,".csv"))) file.remove(paste0("sim",object@tag,"c",i,".csv"))
} else {
if(file.exists(paste0(object@simfile,"c",i,".csv"))) file.remove(paste0(object@simfile,"c",i,".csv"))
}
}
} else {
for(isim in c(1:object@simNo)){
if(object@iCopula==TRUE) {
freqs <- copulaSample(object@freqCopula,nmonths)
}
message(paste0("Simulation ",isim+object@startNo-1," started at ",date()))
for(l in object@lines){
for(t in object@types){
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if(co@line==l & co@claimType==t){
co@simno<-as.character(isim+object@startNo-1)
if(object@iIBNR==TRUE){
if(object@iCopula==TRUE){
co@IBNRfreqIndex@startDate <- startDate
lid <- match(l,object@lines)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),startDate,evaluationDate,co@reportLag,TRUE))
} else {
co@IBNRfreqIndex@startDate <- startDate
mthfreq <- doSample(co@frequency,nmonths)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(mthfreq,startDate,evaluationDate,co@reportLag,TRUE))
}
}
if(object@iUPR==TRUE){
if(object@iCopula==TRUE){
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
lid <- match(l,object@lines)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
} else {
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
mthfreq <- doSample(co@frequency,nmonths)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(mthfreq,as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
}
}
stmp = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
Paid=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
co@iRBNER = object@iRBNER
co@iROPEN = object@iROPEN
co@iIBNR = object@iIBNR
co@iUPR = object@iUPR
stmp<-rbind(stmp,claimSample(co,claimData,startDate,evaluationDate)) #futureDate
if(file.exists(paste0(direc,"/",filename))){
write.table(stmp,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = FALSE)
}else{
write.table(stmp,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = TRUE)
}
simdata <- rbind(simdata,stmp)
# if (isim %% 10 == 0) {
# message(paste0("Simulation ",isim,": ",proc.time()[3]," elapsed"))
}
}
}
#write.csv(simdata,filename,row.names=FALSE)
message(paste0("Simulation ",isim+object@startNo-1," ended at ",date()))
message(">>>Simulation is finished successfully<<<")
}
# write.csv(simdata,filename,row.names=FALSE)
#file.rename("simdatatmp.csv", filename)
}
gc()
return(simdata)
} else {
stop("Nothing to simulate.")
}
}, error = function(err){
if (exists("l") & exists("t") & exists("c") & exists("isim")){
message(paste0(">>>Critical Error for claim simulation: ",err," Line-", l, " Type-", t, " Class-", c, " Sim No.-", isim+object@startNo-1))
} else {
message(paste0(">>>Critical Error for claim simulation: ",err))
}
message(">>>Simulation is finished with error")
gc()
return(-1)
})
})
#' Claim simulation result summary
#' @name simSummary
#' @param object Simulation object
#' @param simdata simulation data generated by claimSimulation
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' #run time is about 30s(>10s) and is commented out here to avoid long waiting time
#' #library(cascsim)
#' #data(claimdata)
#' #lines <- c("Auto")
#' #types <- c("N")
#' #AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' #AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' #AutoN@frequency <- new("Poisson", p1 =50)
#' #AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' #AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' #AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' #AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' #AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' #AutoN@reportLag <- new("Exponential", p1 =0.1)
#' #AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' #AutoN@iCopula <- TRUE #use copula
#' #AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' #param = c(0.1,0.2,0.1))#A Gaussian Copula
#' #AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' #AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' #paras =c(5,1.5,0.005,1.2,3))
#' #AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' #meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' #AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' #meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' #AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' #meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' #AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' #AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' #simobj <- new("Simulation", lines=lines, types=types,
#' #claimobjs= list(AutoN),workingFolder=tempdir())
#' #simobj@simNo <- 1
#' #simobj@iRBNER <-FALSE
#' #simobj@iROPEN <-FALSE
#' #simobj@iIBNR <-TRUE
#' #simobj@iUPR <-FALSE
#' #simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' #evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' #simSummary <- simSummary(simobj,simdata, startDate = as.Date("2012-01-01"))
#' @rdname simSummary-methods
#' @exportMethod simSummary
setGeneric("simSummary", function(object, simdata, ...) standardGeneric("simSummary"))
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims).
#' @rdname simSummary-methods
#' @aliases simSummary,ANY-method
setMethod("simSummary", signature("Simulation", "data.frame"), function(object, simdata, startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"))
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
if(object@sumfile==""){
summaryname <- paste0("sum",object@tag,".csv")
} else {
summaryname <- paste0(object@sumfile,".csv")
}
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if(nrow(simdata)==0) {stop("No simulation data to summarize")}
if (object@iSummary==TRUE){
message(paste0("Starting summarizing the simulation results: ",date()))
# if(length(object@lines)>1) {
idx1<-c(object@lines,"Total")
# } else {
# idx1<-object@lines
# }
# if(length(object@types)>1) {
idx2<-c(object@types,"Total")
# } else {
# idx2<-object@types
# }
# if(length(object@classes)>1) {
idx3 <- vector()
if (object@iRBNER==TRUE) {idx3 <- c(idx3,"IBNER")}
if (object@iROPEN==TRUE) {idx3 <- c(idx3,"ROPEN")}
if (object@iIBNR==TRUE) {idx3 <- c(idx3,"IBNR")}
if (object@iUPR==TRUE) {idx3 <- c(idx3,"UPR")}
idx3<-c(idx3,"Total")
# } else {
# idx3<-object@classes
# }
idx4<-c(as.numeric(substr(as.character(startDate),1,4)):as.numeric(substr(as.character(futureDate),1,4)))
idx4<-c(idx4,"Total")
idx5<-c("avg","sd","min",paste0("percentile ",c(10, 30, 50, 75, 80, 95, 99)),"max")
simSummary<-vector()
summaryNames <- c("LoB","Type","Class","AccidentYear","Measure","Frequency","AvgIncurredLoss","TotalIncurredLoss","AvgUltimateLoss","TotalUltimateLoss","TotalPaid","TotalLAE","PaidLAE","ReopenProb")
simdata$occurrenceYear <- as.numeric(substr(as.character(as.Date(simdata$occurrenceDate)),1,4))
transStatus<-function(class){
if(class=="IBNER"){
return("OPEN")
} else if (class=="ROPEN") {
return("CLOSED")
} else {
return(class)
}
}
for(i1 in idx1){
l<-i1
for(i2 in idx2){
t<-i2
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if((co@line==l & co@claimType==t) | (i1 == "Total") | (i2 == "Total")){
for(i3 in idx3){
c<-i3
#isim<--10000
for(i4 in idx4){
if(i1 != "Total"){
datatmp <- simdata[simdata[,"LoB"] == i1,]
if(i2 != "Total") {datatmp <- datatmp[datatmp[,"Type"] == i2,]}
if(i3 != "Total") {datatmp <- datatmp[datatmp[,"status"] == transStatus(i3),]}
if(i3 == "ROPEN") {
closeCount <- nrow(datatmp)/length(unique(datatmp$Sim))
datatmp <- datatmp[!is.na(datatmp$reopenLoss),]
}
# if(i3 == "Total") {
# datatmp <- datatmp[!(datatmp$status==transStatus("ROPEN") & is.na(datatmp$reopenLoss)),]
# }
if(i4 != "Total") {datatmp <- datatmp[datatmp[,"occurrenceYear"] == as.numeric(i4),]}
}else{
datatmp <- simdata
if(i2 != "Total") {datatmp <- datatmp[datatmp[,"Type"] == i2,]}
if(i3 != "Total") {datatmp <- datatmp[datatmp[,"status"] == transStatus(i3),]}
if(i3 == "ROPEN") {
closeCount <- nrow(datatmp)/length(unique(datatmp$Sim))
datatmp <- datatmp[!is.na(datatmp$reopenLoss),]
}
if(i4 != "Total") {datatmp <- datatmp[datatmp[,"occurrenceYear"] == as.numeric(i4),]}
#datatmp <- datatmp[!(datatmp$status==transStatus("ROPEN") & is.na(datatmp$reopenLoss)),]
}
if (nrow(datatmp)>0){
colreserve <- c("Sim","status","incurredLoss","osRatio","ultimateLoss","reopenLoss","LAE","ultimateLAE")
datatmp <- datatmp[,colreserve]
datatmp$paid <- datatmp$incurredLoss*(1-datatmp$osRatio)
datatmp$ultimateLoss <- ifelse(datatmp$status=="CLOSED" & !is.na(datatmp$reopenLoss),datatmp$reopenLoss,datatmp$ultimateLoss)
datatmp$ultimateLAE <- ifelse(datatmp$status=="CLOSED" & is.na(datatmp$reopenLoss),datatmp$LAE,datatmp$ultimateLAE)
datatmp$status<-NULL
datatmp$reopenLoss<-NULL
frequency<-aggregate(incurredLoss ~ Sim, data = datatmp, length)[,2]
avgIncurredLoss<-aggregate(incurredLoss ~ Sim, data = datatmp, mean)[,2]
totalIncurredLoss<-aggregate(incurredLoss ~ Sim, data = datatmp, sum)[,2]
avgUltimateLoss<-aggregate(ultimateLoss ~ Sim, data = datatmp, mean)[,2]
totalUltimateLoss<-aggregate(ultimateLoss ~ Sim, data = datatmp, sum)[,2]
totalPaidLoss<-aggregate(paid ~ Sim, data = datatmp, sum)[,2]
totalLAE<-aggregate(ultimateLAE ~ Sim, data = datatmp, sum)[,2]
totalPaidLAE<-aggregate(LAE ~ Sim, data = datatmp, sum)[,2]
if(i3=="ROPEN") reopenProb<-aggregate(incurredLoss ~ Sim, data = datatmp, length)[,2]/closeCount
gc()
for(i5 in idx5){
if(i5 == "avg"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,mean(frequency),mean(avgIncurredLoss),mean(totalIncurredLoss),mean(avgUltimateLoss),mean(totalUltimateLoss),mean(totalPaidLoss),mean(totalLAE),mean(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,mean(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else if(i5 == "sd"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,sd(frequency),sd(avgIncurredLoss),sd(totalIncurredLoss),sd(avgUltimateLoss),sd(totalUltimateLoss),sd(totalPaidLoss),sd(totalLAE),sd(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,sd(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else if(i5 == "min"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,min(frequency),min(avgIncurredLoss),min(totalIncurredLoss),min(avgUltimateLoss),min(totalUltimateLoss),min(totalPaidLoss),min(totalLAE),min(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,min(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else if(i5 == "max"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,max(frequency),max(avgIncurredLoss),max(totalIncurredLoss),max(avgUltimateLoss),max(totalUltimateLoss),max(totalPaidLoss),max(totalLAE),max(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,max(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else {
pc <- as.numeric(substr(i5,12,nchar(i5)))/100
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,quantile(frequency,pc),quantile(avgIncurredLoss,pc),quantile(totalIncurredLoss,pc),quantile(avgUltimateLoss,pc),quantile(totalUltimateLoss,pc),quantile(totalPaidLoss,pc),quantile(totalLAE,pc),quantile(totalPaidLAE,pc))
if(i3=="ROPEN"){
tl <- c(tl,quantile(reopenProb,pc))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
}
simSummary<-rbind(simSummary,tl)
}
}
}
}
}
}
message(paste0("Business Line ",i1," summarized: ",date()))
}
rownames(simSummary)<-NULL
simSummary<-as.data.frame(simSummary)
#simSummary$AccidentYear=round(as.numeric(as.character(simSummary$AccidentYear)))
simSummary$Frequency=round(as.numeric(as.character(simSummary$Frequency)))
simSummary$AvgIncurredLoss=round(as.numeric(as.character(simSummary$AvgIncurredLoss)),2)
simSummary$TotalIncurredLoss=round(as.numeric(as.character(simSummary$TotalIncurredLoss)),2)
simSummary$AvgUltimateLoss=round(as.numeric(as.character(simSummary$AvgUltimateLoss)),2)
simSummary$TotalUltimateLoss=round(as.numeric(as.character(simSummary$TotalUltimateLoss)),2)
simSummary$TotalPaid=round(as.numeric(as.character(simSummary$TotalPaid)),2)
simSummary$TotalLAE=round(as.numeric(as.character(simSummary$TotalLAE)),2)
simSummary$PaidLAE=round(as.numeric(as.character(simSummary$PaidLAE)),2)
simSummary$ReopenProb <- ifelse(is.na(simSummary$ReopenProb),NA,round(as.numeric(as.character(simSummary$ReopenProb)),4))
write.table(simSummary,paste0(direc,"/",summaryname),row.names=FALSE, sep=",")
return(simSummary)
} else {
stop("Reset iSummary of the simulation object to TRUE to start summarizing the result.")
}
}, error = function(err){
if (exists("l") & exists("t") & exists("c")){
message(paste0(">>>Critical Error for claim simulation summarization: ",err," Line-", l, " Type-", t, " Class-", c))
} else {
message(paste0(">>>Critical Error for claim simulation summarization: ",err))
}
return(-1)
gc()
})
})
#' Claim simulation result triangles
#' @name simTriangle
#' @param object Simulation object
#' @param claimdata claim data used as basis for simulation
#' @param simdata simulation data generated by claimSimulation
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' #run time is about 30s(>10s) and is commented out here to avoid long waiting time
#' #library(cascsim)
#' #data(claimdata)
#' #lines <- c("Auto")
#' #types <- c("N")
#' #AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' #AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' #AutoN@frequency <- new("Poisson", p1 =50)
#' #AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' #AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' #AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' #AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' #AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' #AutoN@reportLag <- new("Exponential", p1 =0.1)
#' #AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' #AutoN@iCopula <- TRUE #use copula
#' #AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' #param = c(0.1,0.2,0.1))#A Gaussian Copula
#' #AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' #AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' #paras =c(5,1.5,0.005,1.2,3))
#' #AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' #meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' #AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' #meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' #AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' #meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' #AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' #AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' #simobj <- new("Simulation", lines=lines, types=types,
#' #claimobjs= list(AutoN),workingFolder=tempdir())
#' #simobj@simNo <- 1
#' #simobj@iRBNER <-FALSE
#' #simobj@iROPEN <-FALSE
#' #simobj@iIBNR <-TRUE
#' #simobj@iUPR <-FALSE
#' #simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' #evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' #simSummary <- simSummary(simobj,simdata, startDate = as.Date("2012-01-01"))
#' #simTriangle <- simTriangle(simobj,claimdata,simdata, startDate = as.Date("2012-01-01"))
#' @rdname simTriangle-methods
#' @importFrom methods new
#' @exportMethod simTriangle
setGeneric("simTriangle", function(object, claimdata, simdata, ...) standardGeneric("simTriangle"))
#' @param frequency triangle frequency, either "yearly" or "quarterly";
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims).
#' @rdname simTriangle-methods
#' @aliases simTriangle,ANY-method
setMethod("simTriangle", signature("Simulation", "data.frame", "data.frame"), function(object, claimdata, simdata, frequency="yearly", startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"))
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
if(object@sumfile==""){
summaryname <- paste0("tri",object@tag,".csv")
} else {
summaryname <- paste0("tri",object@sumfile,".csv")
}
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if (nrow(claimdata)>0){
claimdata <- claimdata[claimdata[,"LoB"] %in% object@lines & claimdata[,"Type"] %in% object@types,]
claimdata[,"status"] = toupper(claimdata[,"status"])
claimdata[,"occurrenceDate"] = toDate(claimdata[,"occurrenceDate"])
claimdata[,"reportDate"] = toDate(claimdata[,"reportDate"])
claimdata[,"settlementDate"] = toDate(claimdata[,"settlementDate"])
}
if (object@iSummary==TRUE){
message(paste0("Starting creating triangles from the simulation results: ",date()))
# if(length(object@lines)>1) {
idx1<-c(object@lines,"Total")
# } else {
# idx1<-object@lines
# }
# if(length(object@types)>1) {
idx2<-c(object@types,"Total")
# } else {
# idx2<-object@types
# }
idx3<-c("reportedCount","closedCount","incurredLoss")
# idx4<-c(as.numeric(substr(as.character(startDate),1,4)):as.numeric(substr(as.character(futureDate),1,4)))
idx5<-c("upper","avg",paste0("percentile ",c(10,50,90,99)))
if (object@simNo < 10) {
idx5<-c("upper","avg")
}
if (nrow(claimdata)==0 & object@simNo >=10) {
idx5<-c("avg",paste0("percentile ",c(10,50,90,99)))
}
if (nrow(claimdata)==0 & object@simNo <10) {
idx5<-c("avg")
}
#
xTri <- new("Triangle", triID = "TRI1", type = "closedCount", startDate=startDate, frequency="yearly", sim=0, iRBNER=object@iRBNER, iROPEN=object@iROPEN)
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,evaluationDate=evaluationDate,futureDate=futureDate)
dynames <- colnames(xTri@rectangle)
simSummary<-vector()
summaryNames <- c("LoB","Type","Triangle","Measure","AccidentPeriod",dynames)
for(i1 in idx1){
l<-i1
for(i2 in idx2){
t<-i2
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if((co@line==l & co@claimType==t) | (i1 == "Total") | (i2 == "Total")){
for(i3 in idx3){
c<-i3
xTri@type <- c
for(i5 in idx5){
if (i5 == "upper"){
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
cname <- colnames(xTri@upper)
rname <- rownames(xTri@upper)
tl<- data.frame(xTri@upper)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else if(i5 == "avg"){
xTri@sim <- 0
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else if(i5 == "min"){
xTri@sim <- NaN
xTri@percentile <- 0
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else if(i5 == "max"){
xTri@sim <- NaN
xTri@percentile <- 100
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else {
xTri@sim <- NaN
xTri@percentile <- as.numeric(substr(i5,12,nchar(i5)))
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
}
tl$LoB <- i1
tl$Type <- i2
tl$Triangle <- i3
tl$Measure <- i5
tl <- tl[,summaryNames]
simSummary<-rbind(simSummary,tl)
}
}
}
}
message(paste0("Business Line ",i1," triangle constructed: ",date()))
}
rownames(simSummary)<-NULL
write.table(simSummary,paste0(direc,"/",summaryname),row.names=FALSE, sep=",")
return(simSummary)
} else {
stop("Reset iSummary of the simulation object to TRUE to start summarizing the triangles.")
}
}, error = function(err){
if (exists("l") & exists("t") & exists("c") & exists("i5")){
message(paste0(">>>Critical Error for constructing claim simulation triangles: ",err," Line-", l, " Type-", t, " Class-", c, " Measure-",i5))
} else {
message(paste0(">>>Critical Error for constructing claim simulation triangles: ",err))
}
return(-1)
gc()
})
})
#' Generate claim simulation result report in html
#' @name simReport
#' @param object ClaimType object
#' @param simSummary simulation result summary generated by simSummary
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' #run time is about 30s(>10s) and is commented out here to avoid long waiting time
#' #library(cascsim)
#' #data(claimdata)
#' #lines <- c("Auto")
#' #types <- c("N")
#' #AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' #AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' #AutoN@frequency <- new("Poisson", p1 =50)
#' #AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' #AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' #AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' #AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' #AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' #AutoN@reportLag <- new("Exponential", p1 =0.1)
#' #AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' #AutoN@iCopula <- TRUE #use copula
#' #AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' #param = c(0.1,0.2,0.1))#A Gaussian Copula
#' #AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' #AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' #paras =c(5,1.5,0.005,1.2,3))
#' #AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' #meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' #AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' #meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' #AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' #meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' #AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' #AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' #simobj <- new("Simulation", lines=lines, types=types,
#' #claimobjs= list(AutoN),workingFolder=tempdir())
#' #simobj@simNo <- 1
#' #simobj@iRBNER <-FALSE
#' #simobj@iROPEN <-FALSE
#' #simobj@iIBNR <-TRUE
#' #simobj@iUPR <-FALSE
#' #simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' #evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' #simSummary <- simSummary(simobj,simdata, startDate = as.Date("2012-01-01"))
#' #simTriangle <- simTriangle(simobj,claimdata,simdata, startDate = as.Date("2016-01-01"))
#' #simReport(simobj, simSummary, simTriangle, startDate = as.Date("2012-01-01"))
#' @rdname simReport-methods
#'
#' @importFrom R2HTML HTML.title HTML HTMLhr HTMLInsertGraph HTMLCSS
#' @import scatterplot3d
#'
#' @exportMethod simReport
setGeneric("simReport", function(object, simSummary, ...) standardGeneric("simReport"))
#' @param simTriangle triangle summary generated by simTriangle;
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims);
#' @param iYear Boolean that indicates whether summary by accident year should be produced in the report.
#' @rdname simReport-methods
#' @aliases simReport,ANY-method
setMethod("simReport", signature("Simulation", "data.frame"), function(object, simSummary, simTriangle=NA, startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"),iYear=FALSE)
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
if (object@iReport==TRUE){
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this simulation object")
}
classes <- vector()
if (object@iRBNER==TRUE) {classes <- c(classes,"IBNER")}
if (object@iROPEN==TRUE) {classes <- c(classes,"REOPEN")}
if (object@iIBNR==TRUE) {classes <- c(classes,"IBNR")}
if (object@iUPR==TRUE) {classes <- c(classes,"UPR")}
#require(R2HTML)
pckgdir <- find.package("cascsim")
pckgdir <- paste0(pckgdir,"/css/")
if (!dir.exists(paste0(direc,"/report"))) {dir.create(paste0(direc,"/report"))}
copied <- file.copy(file.path(pckgdir, "report.css"), file.path(paste0(direc,"/report/"),"report.css"))
if(object@htmlRpt==""){
htmlfile <- paste0("sim",object@tag)
} else {
htmlfile <- paste0(object@htmlRpt)
}
if (file.exists(paste0(direc,"/report/",htmlfile,".html"))) {file.remove(paste0(direc,"/report/",htmlfile,".html"))}
ReportBegin <- function(file, title, directory = direc) {
file = file.path(paste0(directory,"/report"), file)
cat(paste("<html><head><title>", title, "</title></head>", "<body bgcolor=#D0D0D0>", sep = ""), file = file, append = FALSE)
}
ReportEnd <- function(file, directory = direc) {
file = file.path(paste0(directory,"/report"), file)
cat("<hr size=1></body></html>", file = file, append = TRUE)
}
haveData<-function(line,type,class){
result <- simSummary[simSummary$LoB==line & simSummary$Type==type & simSummary$Class==class,]
if (nrow(result) > 0) {
result <- TRUE
} else {
result <- FALSE
}
return (result)
}
RptFetch<-function(line,type,class,accYear=TRUE){
result <- simSummary[simSummary$LoB==line & simSummary$Type==type & simSummary$Class==class,]
if (nrow(result) > 0) {
rowR<-c("avg","sd","min","percentile 10","percentile 30","percentile 50","percentile 75","percentile 90","percentile 95","percentile 99","max")#"percentile 99.5"
result <- subset(result, result$Measure %in% rowR)
if (accYear==FALSE) {result <- result[result$AccidentYear=="Total",]}
rownames(result)<-NULL
if(accYear==TRUE){
result <- result[,colnames(result) %in% c("AccidentYear","Measure","Frequency","TotalUltimateLoss")]
yrs <- unique(result$AccidentYear)
seps <- list()
coln <- vector()
for(y in c(1:length(yrs))){
tmp <- result[result$AccidentYear==yrs[y],]
if(y==1){
tmp <- tmp[,colnames(tmp) %in% c("Measure","Frequency","TotalUltimateLoss")]
coln <- c(coln,"Measure",paste0(yrs[y],"Count"),paste0(yrs[y],"Loss"))
seps<-c(seps,tmp)
} else {
tmp <- tmp[,colnames(tmp) %in% c("Frequency","TotalUltimateLoss")]
coln <- c(coln,paste0(yrs[y],"Count"),paste0(yrs[y],"Loss"))
seps<-c(seps,tmp)
}
}
for(y in c(1:length(seps))){
if(y==1){
result<-seps[[y]]
}else{
result<-cbind(result,seps[[y]])
}
}
result<-as.data.frame(result)
result[,1]<-rowR
colnames(result)<-coln
} else {
result <- result[,!colnames(result) %in% c("LoB","Type","Class","AccidentYear","AvgIncurredLoss","TotalIncurredLoss","TotalPaid","PaidLAE","ReopenProb")]
colnames(result) <- c("Measure","Count","AvgUltimateLoss","TotalUltimateLoss","TotalLAE")
}
} else {
result <- paste0("No ", class, " claims are simulated.")
}
return(result)
}
triangleFetch<-function(line,type,triangle,measure){
result <- simTriangle[simTriangle$LoB==line & simTriangle$Type==type & simTriangle$Triangle==triangle & simTriangle$Measure==measure,]
if (nrow(result) > 0) {
colremove <- c("LoB","Type","Triangle","Measure")
colkeep <- colnames(result)
colkeep <- colkeep[!colkeep %in% colremove]
if (measure == "upper") {colkeep <- colkeep[1:(nrow(result)+1)]}
result <- result[,colnames(result) %in% colkeep]
if (triangle=="incurredLoss" | triangle=="paidLoss"){
result[,!colnames(result) %in% c("AccidentPeriod")] <- round(result[,!colnames(result) %in% c("AccidentPeriod")]/1000)
}
for (i in c(2:ncol(result))){
result[,i] <- formatC(result[,i], format="d", big.mark=",")
}
if (measure == "upper") {
for (c in range(1:ncol(result))){
result[,c] = as.character(result[,c])
}
#result <- apply(result,2,as.character)
for(i in c(1:nrow(result))){
for(j in c(1:ncol(result))){
if(result[i,j]=="0"){result[i,j]=""}
}
}
} else {
result$AccidentPeriod <- as.character(result$AccidentPeriod)
}
colnames(result) <- colkeep
rownames(result) <- NULL
} else {
result <- NULL
}
return(result)
}
RptPlot<-function(line,type,class,accYear){
result <- simSummary[simSummary$LoB==line & simSummary$Type==type & simSummary$Class==class & simSummary$AccidentYear==accYear,]
##c("avg","sd","min",paste0("percentile ",c(10, 30, 50, 75, 80, 95, 99)),"max")
rowR<-paste0("percentile ",c(10, 30, 50, 75, 80, 95, 99))
result <- subset(result, result$Measure %in% rowR)
perc <- c(10, 30, 50, 75, 80, 95, 99)/100
par(mfrow = c(2, 2))
plot(result$TotalUltimateLoss, perc, type = "p", xlab = "TotalUltimateLoss", ylab = "Cumulative Probability", main = "Total Ultimate Loss CDF")
plot(result$AvgUltimateLoss, perc, type = "p", xlab = "AvgUltimateLoss", ylab = "Cumulative Probability", main = "Avg. Ultimate Loss CDF")
plot(result$Frequency, perc, type = "p", xlab = "Frequency", ylab = "Cumulative Probability", main = "Count CDF")
plot(result$TotalLAE, perc, type = "p", xlab = "TotalLAE", ylab = "Cumulative Probability", main = "Total LAE CDF")
}
distPlot<-function(claimobj){
par(mfrow = c(2, 2))
x<-doSample(claimobj@frequency, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@frequency, p)
plot(q, Density(claimobj@frequency, q),type="l", main = "Frequency", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@severity, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@severity, p)
plot(q, Density(claimobj@severity, q),type="l", main = "Severity", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@reportLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@reportLag, p)
plot(q, Density(claimobj@reportLag, q),type="l", main = "Report Lag", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@settlementLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@settlementLag, p)
plot(q, Density(claimobj@settlementLag, q),type="l", main = "Settlement Lag", xlab="q", ylab="density", col = "red")
}
idxPlot<-function(claimobj){
par(mfrow = c(2, 2))
plot(claimobj@exposureIndex@monthlyIndex, main = "Exposure Index", xlab="Time", ylab="Index")
plot(claimobj@severityIndex@monthlyIndex, main = "Severity Index", xlab="Time", ylab="Index")
if(claimobj@iCopula==TRUE){
samples<-copulaSample(claimobj@ssrCopula, 1000)
if (claimobj@ssrCopula@dimension==2)
{
plot(samples, col="blue")
}
else if (claimobj@ssrCopula@dimension>=2)
{
#require(scatterplot3d)
scatterplot3d(samples[,1:3], color="blue")
}
} else {
plotText("Severity and lags are independent.")
}
if (claimobj@laeDevFac@FacModel==TRUE){
plotText("LAE development factor is formula based.")
} else {
plot(claimobj@laeDevFac@meanList, main = "LAE Development", xlab="Duration", ylab="Factor")
}
}
roPlot<-function(claimobj){
par(mfrow = c(2, 2))
if (claimobj@reopen@FacModel==TRUE){
plotText("Reopen probability is formula based.")
} else {
plot(claimobj@reopen@meanList, main = "Reopen probability", xlab="Duration", ylab="Probability")
}
if (claimobj@roDevFac@FacModel==TRUE){
plotText("Reopen loss development is formula based.")
} else if (claimobj@irDevFac==3 & claimobj@roDevFac@distType == "normal"){
plot(claimobj@roDevFac@meanList, main = "Reopen loss development", xlab="Duration", ylab="Factor")
} else if (claimobj@irDevFac==3 & claimobj@roDevFac@distType != "normal"){
plotText(paste0("LDF based on ", claimobj@roDevFac@distType, " dist."))
} else {
plotText("Loss based on truncated severity dist.")
}
x<-doSample(claimobj@reopenLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@reopenLag, p)
plot(q, Density(claimobj@reopenLag, q),type="l", main = "Reopen Lag", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@resettleLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@resettleLag, p)
plot(q, Density(claimobj@resettleLag, q),type="l", main = "Resettlement Lag", xlab="q", ylab="density", col = "red")
}
dlPlot<-function(claimobj){
par(mfrow = c(1, 2))
y<-claimobj@deductible
ysample<-doSample(y, 1000)
d1<-ecdf(ysample)
plot(d1, main = "CDF: Deductible", col="blue")
y<-claimobj@limit
ysample<-doSample(y, 1000)
d1<-ecdf(ysample)
plot(d1, main = "CDF: Limit", col="blue")
}
fibnerPlot<-function(claimobj){
par(mfrow = c(1, 1))
if (claimobj@fIBNER@FacModel==TRUE){
plotText("IBNER loss development is formula based.")
} else {
plot(claimobj@fIBNER@meanList, main = "RBNER Loss Development", xlab="Duration", ylab="Factor")
}
}
p0Plot<-function(claimobj){
par(mfrow = c(1, 1))
plot(claimobj@p0@meanList, main = "Probability of Zero Payment", xlab="Duration", ylab="Factor")
}
vecFlatten<-function(vec){
result<-""
if(length(vec)>1){
for(i in c(1:(length(vec)-1))){
result<-paste0(result,vec[i],", ")
}
result <- paste0(result,"and ",vec[length(vec)])
} else {
result <- vec[1]
}
return(result)
}
getCopulaInfo<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-object@lines
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
getCopulaSR<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-c("Severity","SettlementLag","ReportLag")
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
findClaimObj <- function(line,type,objpool){
for(co in objpool){
if(co@line==line & co@claimType==type) {
result <- co
break
} else {
result <- NULL
}
}
return(result)
}
HTMLcontent <- function(file, append = TRUE, directory = direc) {
file = file.path(paste0(directory,"/report"), file)
cat("\n", file = file, append = append)
HTML.title("Claim Simulation Report", file=file, HR=1)
HTML(paste0("Claim Data Start Date: ",startDate,"; Claim Data Ending Date: ",evaluationDate,"; Future Claim Ending Date: ",futureDate), file=file)
HTML(paste0("Simulation Task ID: ",object@tag), file=file)
HTML(paste0("Report generated at ",date()), file=file)
HTMLhr(file = file)
HTML.title("Notation",file = file, HR=4)
HTML(paste0("AccidentPeriod: Accident Year or Quarter."),file=file)
HTML(paste0("Mxx: Development period. For example, M12 is the first development year and M24 is the second development year."),file=file)
HTML(paste0("Count: Number of simulated claims."),file=file)
HTML(paste0("AvgUltimateLoss: Ultimate Loss per claim. It is the average loss of all claims in one portfolio simulation. Each simulation has an average ultimate loss and multiple simulations allow the construction of the distribution."),file=file)
HTML(paste0("TotalUltimateLoss: Total Ultimate Loss for all simulated claims in one portfolio simulation. Multiple simulations allow the construction of its distribution."),file=file)
HTML(paste0("TotalLAE: Total Loss Adjustment Expense for all simulated claims in one portfolio simulation(if modeled separately from indemnity). Multiple simulations allow the construction of its distribution."),file=file)
HTML(paste0("Measure: Statistical measure such as avg(average), sd(standard deviation), min, max and a certain percentile."),file=file)
HTMLhr(file = file)
HTML.title("Links",file = file, HR=4)
HTML(paste0("<a href=\"#tp\">Total Portfolio</a>"),file=file)
for(l in object@lines){
HTML(paste0("<a href=\"#",l,"\">",l,"-Total","</a>"),file=file)
for(t in object@types){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
s<-paste0("<a href=\"#",l,t,"\">",l,"-",t,"</a>")
for(c in classes){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
s<-paste0(s," ","<a href=\"#",l,t,c,"\">",c,"</a>")
}
}
HTML(s,file=file)
}
}
}
HTMLhr(file = file)
HTML(paste0("<a name=\"tp\"></a>"), file = file)
HTML.title("Total Portfolio",file = file, HR=2)
HTML(paste0("The portfolio includes ",length(object@lines)," business line(s): ",vecFlatten(object@lines),". Each business line contains ",length(object@types)," type(s): ",vecFlatten(object@types),". ",length(classes)," class(es) of claims are simulated: ",vecFlatten(classes),"."),file=file)
if(!(is.na(simTriangle)[1])){
if (is.null(triangleFetch("Total","Total","reportedCount","upper"))) {
HTML.title("Average Simulated Reported Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","reportedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Average Simulated Closed Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","closedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Average Simulated Data Incurred Loss Rectangle (Thousands)",file = file, HR=3)
HTML(triangleFetch("Total","Total","incurredLoss","avg"), file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML.title("Claim Data Reported Count Triangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","reportedCount","upper"), file = file, innerBorder=0.5, row.names=FALSE)
HTML.title("Average Simulated Reported Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","reportedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Claim Data Closed Count Triangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","closedCount","upper"), file = file, innerBorder=0.5, row.names=FALSE)
HTML.title("Average Simulated Closed Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","closedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Claim Data Incurred Loss Triangle (Thousands)",file = file, HR=3)
HTML(triangleFetch("Total","Total","incurredLoss","upper"), file = file, innerBorder=0.5, row.names=FALSE)
HTML.title("Average Simulated Data Incurred Loss Rectangle (Thousands)",file = file, HR=3)
HTML(triangleFetch("Total","Total","incurredLoss","avg"), file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLhr(file = file)
}
HTML.title("Key Statistics - Total Portfolio, All Accident Years",file = file, HR=3)
HTML(RptFetch("Total","Total","Total",FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("CDF - Total Portfolio, All Accident Years",file = file, HR=3)
png(filename = paste0(direc,"/report/","sim",object@tag,"tp.png"))
RptPlot("Total","Total","Total","Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,"tp.png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title("Key Statistics - Total Portfolio, By Accident Year",file = file, HR=3)
HTML(RptFetch("Total","Total","Total",TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
HTML.title("Annual Frequency Correlation",file = file, HR=3)
if(object@iCopula==FALSE){
HTML("Frequencies are assumed to be independent for different lines/types", file = file)
} else {
HTML(paste0("Frequencies are assumed to follow ",getCopulaInfo(object@freqCopula)$out), file = file)
HTML(getCopulaInfo(object@freqCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLhr(file = file)
for(l in object@lines){
HTML(paste0("<a name=\"",l,"\"></a>"), file = file)
HTML.title(l,file = file, HR=2)
HTML(RptFetch(l,"Total","Total",FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title(paste0("CDF - ",l,", All Accident Years"),file = file, HR=3)
png(filename = paste0(direc,"/report/","sim",object@tag,l,".png"))
RptPlot(l,"Total","Total","Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,".png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title(paste0("Key Statistics - ",l,", By Accident Year"),file = file, HR=3)
HTML(RptFetch(l,"Total","Total",TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
for(t in object@types){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("<a name=\"",l,t,"\"></a>"), file = file)
HTML.title(paste0(l,"-",t),file = file, HR=2)
HTML(RptFetch(l,t,"Total",FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title(paste0("CDF - ",l,"-",t,", All Accident Years"),file = file, HR=3)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,".png"))
RptPlot(l,t,"Total","Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,".png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title(paste0("Key Statistics - ",l,"-",t,", By Accident Year"),file = file, HR=3)
HTML(RptFetch(l,t,"Total",TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Probability of Zero Payment by Development Year(p0): ",findClaimObj(l,t,object@claimobjs)@p0@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@p0), file = file, innerBorder=0.5, row.names=TRUE)
HTMLhr(file = file)
}
for(c in classes){
HTML(paste0("<a name=\"",l,t,c,"\"></a>"), file = file)
HTML.title(paste0(l,"-",t,"-",c),file = file, HR=3)
if (c=="IBNER"){
c2 = "IBNER"
} else if (c=="REOPEN") {
c2 = "ROPEN"
} else {
c2 = c
}
HTML(RptFetch(l,t,c2,FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
if (haveData(l,t,c2) == TRUE) {
HTML.title(paste0("CDF - ",l,"-",t,"-",c,", All Accident Years"),file = file, HR=5)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,".png"))
RptPlot(l,t,c2,"Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,".png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title(paste0("Key Statistics - ",l,"-",t,"-",c,", By Accident Year"),file = file, HR=5)
HTML(RptFetch(l,t,c,TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
HTML.title("Assumptions",file = file, HR=5)
if(c=="IBNER"){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
if(findClaimObj(l,t,object@claimobjs)@fIBNER@FacModel==TRUE & findClaimObj(l,t,object@claimobjs)@ioDevFac==3){
HTML(paste0("IBNER Loss Development: ",findClaimObj(l,t,object@claimobjs)@fIBNER@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@fIBNER)),file=file)
} else if (findClaimObj(l,t,object@claimobjs)@ioDevFac==3){
HTML(paste0("IBNER Loss Development: ",findClaimObj(l,t,object@claimobjs)@fIBNER@FacID, ". Distribution type: ", findClaimObj(l,t,object@claimobjs)@fIBNER@distType),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@fIBNER), file = file, innerBorder=0.5, row.names=TRUE)
} else if (findClaimObj(l,t,object@claimobjs)@ioDevFac==2){
HTML(paste0("IBNER Loss Development: Severity distribution floored by incurred loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
} else {
HTML(paste0("IBNER Loss Development: Severity distribution conditioned on paid loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
}
if (findClaimObj(l,t,object@claimobjs)@ioDevFac==3 & findClaimObj(l,t,object@claimobjs)@fIBNER@distType == "normal"){
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"fibner.png"),width=240,height=240)
fibnerPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"fibner.png"),Caption = "",file = file,WidthHTML=250)
}
}
}else if (c=="REOPEN"){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
if(findClaimObj(l,t,object@claimobjs)@reopen@FacModel==TRUE){
HTML(paste0("Reopen Probability: ",findClaimObj(l,t,object@claimobjs)@reopen@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@reopen)),file=file)
} else {
HTML(paste0("Reopen Probability: ",findClaimObj(l,t,object@claimobjs)@reopen@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@reopen), file = file, innerBorder=0.5, row.names=TRUE)
}
if(findClaimObj(l,t,object@claimobjs)@roDevFac@FacModel==TRUE & findClaimObj(l,t,object@claimobjs)@irDevFac==3){
HTML(paste0("Reopen Loss Development: ",findClaimObj(l,t,object@claimobjs)@roDevFac@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@roDevFac)),file=file)
} else if (findClaimObj(l,t,object@claimobjs)@irDevFac==3){
HTML(paste0("Reopen Loss Development: ",findClaimObj(l,t,object@claimobjs)@roDevFac@FacID, ". Distribution type: ", findClaimObj(l,t,object@claimobjs)@roDevFac@distType),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@roDevFac), file = file, innerBorder=0.5, row.names=TRUE)
} else if (findClaimObj(l,t,object@claimobjs)@irDevFac==2){
HTML(paste0("Reopen Loss Development: Severity distribution floored by incurred loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
} else {
HTML(paste0("Reopen Loss Development: Severity distribution conditioned on paid loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
}
HTML(paste0("Reopen Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reopenLag)),file=file)
HTML(paste0("Resettlement Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@resettleLag)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"ro.png"))
roPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"ro.png"),Caption = "",file = file)
}
}else if (c=="IBNR"){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Frequency Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@frequency)),file=file)
HTML(paste0("Severity Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
HTML(paste0("Report Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reportLag)),file=file)
HTML(paste0("Settlement Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@settlementLag)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"))
distPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"),Caption = "",file = file)
HTML(paste0("Severity Index: ",toString(findClaimObj(l,t,object@claimobjs)@severityIndex)),file=file)
HTML(paste0("Exposure Index: ",toString(findClaimObj(l,t,object@claimobjs)@exposureIndex)),file=file)
HTML(paste0("Correlation between Severity, settlement Lag and Report Lag: "),file=file)
if(findClaimObj(l,t,object@claimobjs)@iCopula==FALSE){
HTML("Severity, Settlement Lag and Report Lag are assumed to be independent.", file = file)
} else {
HTML(paste0("Severity, Settlement Lag and Report Lag are assumed to follow ",getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$out), file = file)
HTML(getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
if(findClaimObj(l,t,object@claimobjs)@laeDevFac@FacModel==TRUE){
HTML(paste0("LAE: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@laeDevFac)),file=file)
} else {
HTML(paste0("LAE Development Schedule: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@laeDevFac), file = file, innerBorder=0.5, row.names=TRUE)
}
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"))
idxPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"),Caption = "",file = file)
HTML(paste0("Deductible Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@deductible)),file=file)
HTML(paste0("Limit Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@limit)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"dl.png"),width=480,height=240)
dlPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"dl.png"),Caption = "",file = file)
}
}else if (c=="UPR" & haveData(l,t,"IBNR")==FALSE){ #sum("IBNR"==classes)==0 &
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Frequency Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@frequency)),file=file)
HTML(paste0("Severity Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
HTML(paste0("Report Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reportLag)),file=file)
HTML(paste0("Settlement Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@settlementLag)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"))
distPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"),Caption = "",file = file)
HTML(paste0("Severity Index: ",toString(findClaimObj(l,t,object@claimobjs)@severityIndex)),file=file)
HTML(paste0("Exposure Index: ",toString(findClaimObj(l,t,object@claimobjs)@exposureIndex)),file=file)
HTML(paste0("Correlation between Severity, settlement Lag and Report Lag: "),file=file)
if(findClaimObj(l,t,object@claimobjs)@iCopula==FALSE){
HTML("Severity, settlement Lag and Report Lag are assumed to be independent.", file = file)
} else {
HTML(paste0("Severity, settlement Lag and Report Lag are assumed to follow ",getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$out), file = file)
HTML(getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
if(findClaimObj(l,t,object@claimobjs)@laeDevFac@FacModel==TRUE){
HTML(paste0("LAE: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@laeDevFac)),file=file)
} else {
HTML(paste0("LAE Development Schedule: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@laeDevFac), file = file, innerBorder=0.5, row.names=TRUE)
}
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"))
idxPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"),Caption = "",file = file)
}
} else {
HTML(paste0("Same as assumptions used for IBNR."),file=file)
}
}
HTMLhr(file = file)
}
}
}
}
message(paste("Report generated: ",file," at ",date(), "\n",sep = ""))
}
Report <- function(file) {
ReportBegin(file,"Claim Simulation Report")
if(copied == TRUE | file.exists(paste0(direc,"/report/","report.css"))) {
HTMLCSS(file = paste0(direc,"/report/",file), CSSfile = paste0(direc,"/report/","report.css"))
}
HTMLcontent(file)
ReportEnd(file)
}
big_mark_opt <- getOption("R2HTML.format.big.mark")
decimal_mark_opt <- getOption("R2HTML.format.decimal.mark")
options("R2HTML.format.big.mark"=",")
options("R2HTML.format.decimal.mark"=".")
Report(paste0(htmlfile,".html"))
on.exit(options("R2HTML.format.big.mark"=big_mark_opt))
on.exit(options("R2HTML.format.decimal.mark"=decimal_mark_opt), add = TRUE)
} else {
stop("Set iReport of simulation object to TRUE to generate html report.")
}
}, error = function(err){
message(paste0(">>>Critical Error for generating claim simulation report: ",err))
gc()
})
})
Try the cascsim package in your browser
Any scripts or data that you put into this service are public.
cascsim documentation built on Jan. 13, 2020, 5:07 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
#' An S4 class to represent a simulation task.
#'
#' @slot startNo The starting simulation index.
#' @slot simNo Number of simulation.
#' @slot lines A string vector to identify the business line(s) to be simulated.
#' @slot types A string vector to identify the claim types to be simulated.
#' @slot iRBNER A Boolean indicating whether IBNER claims need to be simulated.
#' @slot iROPEN A Boolean indicating whether claim reopening needs to be simulated.
#' @slot iIBNR A Boolean indicating whether IBNR claims need to be simulated.
#' @slot iUPR A Boolean indicating whether future claims need to be simulated.
#' @slot claimobjs A list of claim objects.
#' @slot workingFolder A string to specify the working folder where the simulation results will be saved.
#' @slot iCopula A Boolean indicating whether to use copula for frequency simulation.
#' @slot freqCopula Frequency copula.
#' @slot iSummary A Boolean indicating whether to summarzie the simulation results.
#' @slot iReport A Boolean indicating whether to generate an HTML report.
#' @slot iFit A Boolean indicating whether to fit some simulation parameters based on claim data.
#' @slot ncores Number of cores used for simulation.
#' @slot tag A unique tag for the simulation object including date and a random ID.
#' @slot fitfile A string to set the distribution fitting file name. If omitted, a name based on tag will be used.
#' @slot copfile A string to set the copula fitting file name. If omitted, a name based on tag will be used.
#' @slot facfile A string to set the factor fitting file name. Factor table is development year dependant. It could be the probability of zero payment, reopen probability, or loss development factors. If omitted, a name based on tag will be used.
#' @slot fitRpt A string to set the distribution fitting html report file name. If omitted, a name based on tag will be used.
#' @slot simfile A string to set the simulation result file name. If omitted, a name based on tag will be used.
#' @slot sumfile A string to set the summary file name. If omitted, a name based on tag will be used.
#' @slot plog A string to set the parallel run log file name. If omitted, a name based on tag will be used.
#' @slot htmlRpt A string to set the html report name. If omitted, a name based on tag will be used.
#' @slot libpath A string to the R liabrary folder where required packages are installed.
setClass("Simulation",
slots=c(
startNo="numeric",
simNo="numeric",
lines="vector",
types="vector",
# classes="vector",
iRBNER="logical",
iROPEN="logical",
iIBNR="logical",
iUPR="logical",
claimobjs="list",
workingFolder="character",
iCopula="logical",
freqCopula="CopulaObj",
iSummary="logical",
iReport="logical",
iFit="logical",
ncores="numeric",
tag="character",
fitfile="character",
copfile="character",
facfile="character",
fitRpt="character",
simfile="character",
sumfile="character",
plog="character",
htmlRpt="character",
libpath="character"
),
prototype=list(
startNo=1,
simNo=1,
lines=vector(),
types=vector(),
# classes=c("RBNER","ROPEN","IBNR","UPR"),
iRBNER=TRUE,
iROPEN=TRUE,
iIBNR=TRUE,
iUPR=TRUE,
claimobjs=list(),
workingFolder="",
iCopula=FALSE,
freqCopula=new("CopulaObj",param=c(0,0,0),dimension=3),
iSummary=TRUE,
iReport=TRUE,
iFit=TRUE,
ncores=1,
tag=paste0(format(Sys.Date(),"%Y-%m-%d"),"-",round(runif(1,1,10000))),
fitfile="",
copfile="",
facfile="",
fitRpt="",
simfile="",
sumfile="",
plog="",
htmlRpt="",
libpath=""
)
)
#' Claim data fitting analysis at line/type/status level
#' @name claimFitting
#' @param object Simulation object
#' @param claimData claim data including existing claims for RBNER and claim reopenness analysis
#' @param ... Additional parameters that may or may not be used.
#' @rdname claimFitting-methods
#'
#' @importFrom methods new
#'
#' @import grDevices
#'
#' @importFrom R2HTML HTML.title HTML HTMLhr HTMLInsertGraph HTMLCSS
#'
#' @exportMethod claimFitting
setGeneric("claimFitting", function(object, claimData, ...) standardGeneric("claimFitting"))
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param lineList List of business lines to be included in claim fitting;
#' @param typeList List of claim types to be included in claim fitting;
#' @param discreteDist List of discrete distributions to try fitting (report lag, settlemet lag, frequency);
#' @param continuousDist List of continuous distribution to try fitting (severity);
#' @param copulaList List of copula to try fitting;
#' @param fReportLag Boolean variable to indicate whether report lag needs to be fitted;
#' @param fSettlementLag Boolean variable to indicate whether settlement lag needs to be fitted;
#' @param fFrequency Boolean variable to indicate whether monthly frequency needs to be fitted;
#' @param fSeverity Boolean variable to indicate whether severity needs to be fitted;
#' @param fSSRCorrelation Boolean variable to indicate whether copula among severity, report lag and settlement lag needs to be fitted;
#' @param fFreqCorrelation Boolean variable to indicate whether copula among frequencies of business lines needs to be fitted.
#' @param copulaTest Whether to test copula. The testing could take a very long time;
#' @param iTotalLoss Boolean variable to indicate whether total loss before deductible and limit is available for severity fitting;
#' @param fDeductible Boolean variable to indicate whether deductible empirical distribution needs to be fitted;
#' @param fLimit Boolean variable to indicate whether limit empirical distribution needs to be fitted;
#' @param check Boolean variable to indicate whether graph of each tried distribution fitting needs to be generated and saved.
#' @examples
#' library(cascsim)
#' data(claimdata)
#' lines<-c("Auto")
#' types<-c("N")
#' #exposure index
#' index1 <- new("Index",monthlyIndex=c(rep(1,11),cumprod(c(1,rep(1.5^(1/12),11))),
#' cumprod(c(1.5,rep((1.3/1.5)^(1/12),11))),
#' cumprod(c(1.3,rep((1.35/1.3)^(1/12),11))),cumprod(c(1.35,rep((1.4/1.35)^(1/12),11))),rep(1.4,301)))
#' #severity index
#' index2 <- new("Index",monthlyIndex=c(cumprod(c(1,rep(1.03^(1/12),59))),rep(1.03^(5),300)))
#' objan <- new("ClaimType", line="Auto",claimType="N",exposureIndex=index1,severityIndex=index2)
#' objlist <- list(objan)
#' simobj <- new("Simulation",lines=lines,types=types,claimobjs=objlist,iFit=TRUE,
#' iCopula=FALSE, iReport=TRUE, workingFolder=tempdir())
#' simobj <- claimFitting(simobj,claimdata,fSSRCorrelation = FALSE, fSettlementLag = FALSE)
#' @rdname claimFitting-methods
#' @aliases claimFitting,ANY-method
setMethod("claimFitting", signature("Simulation", "data.frame"), function(object, claimData, startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),
lineList = object@lines,
typeList = object@types,
discreteDist = c("Poisson","NegativeBinomial","Geometric"),
continuousDist = c("Normal","Lognormal","Pareto","Weibull","Gamma","Uniform","Exponential"),
copulaList = c("normal"),#c("normal","clayton","gumbel","frank","joe", "t")
fReportLag = TRUE,
fSettlementLag = TRUE,
fFrequency = TRUE,
fSeverity = TRUE,
fSSRCorrelation = TRUE,
fFreqCorrelation = TRUE,
copulaTest = TRUE,
iTotalLoss = TRUE,
fDeductible = TRUE,
fLimit = TRUE,
check = TRUE)
{
tryCatch({
message("Fitting process started.")
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
if (nrow(claimData) > 0) {
claimData[,"occurrenceDate"] = toDate(claimData[,"occurrenceDate"])
claimData[,"reportDate"] = toDate(claimData[,"reportDate"])
claimData[,"settlementDate"] = toDate(claimData[,"settlementDate"])
claimData[,"status"] = toupper(claimData[,"status"])
claimData[,"claimLiability"] = toupper(claimData[,"claimLiability"])
claimData <- claimData[as.Date(claimData[,"occurrenceDate"])>=as.Date(startDate),]
claimData <- claimData[as.Date(claimData[,"reportDate"])<=as.Date(evaluationDate),]
}
if (as.numeric(as.Date(evaluationDate) - as.Date(startDate))<60 & fFrequency == TRUE) {
fFrequency = FALSE
fFreqCorrelation = FALSE
message("Frequency and Frequency Copula Fitting are turned off due to insufficient data.")
message("\n")
}
if(length(object@lines)>0 & length(object@types)>0 & length(object@claimobjs)>0 & nrow(claimData)>0) {
if(object@iFit==TRUE){
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if(object@iReport==TRUE){
if (!dir.exists(paste0(direc,"/fit"))) {dir.create(paste0(direc,"/fit"))}
}
if(object@fitfile==""){
fitname <- paste0("fit",object@tag,".csv")
} else {
fitname <- paste0(object@fitfile,".csv")
}
if(object@fitfile==""){
empname <- paste0("emp",object@tag,".csv")
} else {
empname <- paste0("emp",object@fitfile,".csv")
}
if(object@copfile==""){
copname <- paste0("cop",object@tag,".csv")
} else {
copname <- paste0(object@copfile,".csv")
}
if(object@facfile==""){
facname <- paste0("fac",object@tag,".csv")
} else {
facname <- paste0(object@facfile,".csv")
}
#f <- c("reportLag", "settlementLag", "frequency", "severity", "ssrCorrelation", "freqCorrelation")
empiricals <- data.frame(matrix(seq(0.001,1,0.001),1000,1))
colnames(empiricals) <- "prob"
fitsummary <- data.frame(LoB=character(),
Type=character(),
Fit=character(),
Distribution=character(),
Method=character(),
Parameter=character(),
SD=character(),
p0=numeric(),
DoF=integer(),
ChiSq=double(),
p=double(),
KS=double(),
pks=double(),
loglik=double(),
AIC=double(),
BIC=double(),
stringsAsFactors=FALSE)
rfit<-1
fitsumcop <- data.frame(LoB=character(),
Type=character(),
Fit=character(),
Copula=character(),
Method=character(),
Parameter=character(),
SD=character(),
DoF=integer(),
Sn=double(),
p=double(),
stringsAsFactors=FALSE)
rcop<-1
fitsumfac <- data.frame(LoB=character(),
Type=character(),
Fit=character(),
Year=character(),
MeanList=double(),
VolList=double(),
stringsAsFactors=FALSE)
rfac<-1
for(l in lineList){#object@lines
for(t in object@types){#object@lines
fitdata <- claimData[claimData[,"LoB"]==l,]
fitdata <- fitdata[fitdata[,"Type"]==t,]
obji<-1
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
obji <- obji + 1
}
if(nrow(fitdata)>5 & co@line==l & co@claimType==t){
if(fReportLag==TRUE){
f <- "reportLag"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
reportlags <- as.numeric(as.Date(fitdata[,"reportDate"])-as.Date(fitdata[,"occurrenceDate"]))
reportlags <- ifelse(reportlags==0, runif(length(reportlags)),reportlags)
if(TRUE){#(objName(co@reportLag)!="Empirical"){
xFit <- new("FitDist", observation=as.data.frame(reportlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
bestBIC <- 1e10
irlso <- 0
for (i in c(1:length(continuousDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(continuousDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
irlso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,continuousDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",continuousDist[i]," distribution"))
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
#message(toString(xFit@fitted))
}
if (irlso == 1) {
co@reportLag <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@reportLag)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@reportLag@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=as.data.frame(reportlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@reportLag <- xFit@fitted
#xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=as.data.frame(reportlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
message("\n")
}
if(fFrequency==TRUE){
f <- "frequency"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
rawdata <- as.data.frame(as.Date(fitdata$occurrenceDate))
colnames(rawdata)<-"occurrenceDate"
if(TRUE){#(objName(co@frequency)!="Empirical"){
xFit <- new("FitDist", observation=rawdata, trend=co@exposureIndex,startDate = startDate, endDate= evaluationDate, method="mle",ifreq=TRUE,idate=TRUE, freq="Monthly", iLag=TRUE, reportLag=co@reportLag)
xFit <- setFitdata(xFit)
bestBIC <- 1e10
ifso <- 0
for (i in c(1:length(discreteDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(discreteDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",discreteDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",discreteDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
ifso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,discreteDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,discreteDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,discreteDist[i],".jpg"))
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",discreteDist[i]," distribution"))
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
#message(toString(xFit@fitted))
}
if (ifso == 1) {
co@frequency <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@frequency)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@frequency@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=rawdata, trend=co@exposureIndex,startDate = startDate, method="mle",ifreq=TRUE,idate=TRUE, freq="Monthly", iLag=TRUE, reportLag=co@reportLag)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@frequency <- xFit@fitted
#xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=rawdata, trend=co@exposureIndex,startDate = startDate, endDate= evaluationDate, method="mle",ifreq=TRUE,idate=TRUE, freq="Monthly", iLag=TRUE, reportLag=co@reportLag)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
message("\n")
}
if(fSettlementLag==TRUE){
fitdatacls <- fitdata[fitdata$status=="CLOSED",]
f <- "settlementLag"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
settlementlags <- as.numeric(as.Date(fitdatacls[,"settlementDate"])-as.Date(fitdatacls[,"reportDate"]))
rm(fitdatacls)
settlementlags <- settlementlags[!is.na(settlementlags)]
settlementlags <- ifelse(settlementlags==0, runif(length(settlementlags)),settlementlags)
if(TRUE){#(objName(co@settlementLag)!="Empirical"){
xFit <- new("FitDist", observation=as.data.frame(settlementlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
bestBIC <- 1e10
islso <- 0
for (i in c(1:length(continuousDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(continuousDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
islso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,continuousDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",continuousDist[i]," distribution"))
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
}
if (islso == 1) {
co@settlementLag <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@settlementLag)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@settlementLag@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=as.data.frame(settlementlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@settlementLag <- xFit@fitted
#xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=as.data.frame(settlementlags), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
message("\n")
}
if(fDeductible==TRUE){
f <- "Deductible"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
deductibles <- as.numeric(fitdata[,"Deductible"])
deductibles <- ifelse(is.na(deductibles),0,deductibles)
deductibles <- ifelse(deductibles<0,0,deductibles)
xFit <- new("FitDist", observation=as.data.frame(deductibles), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
co@deductible <- xFit@fitted
rfit <- rfit + 1
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=240,height=240)
CDFPlot(xFit)
dev.off()
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical",".jpg"))
par(mfrow = c(1, 1))
CDFPlot(xFit)
dev.off()
}
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
message("\n")
}
if(fLimit==TRUE){
f <- "Limit"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
limits <- as.numeric(fitdata[,"Limit"])
limits <- ifelse(is.na(limits),1e10,limits)
limits <- ifelse(limits<0,1e10,limits)
xFit <- new("FitDist", observation=as.data.frame(limits), method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- NA
co@limit <- xFit@fitted
rfit <- rfit + 1
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=240,height=240)
CDFPlot(xFit)
dev.off()
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical",".jpg"))
par(mfrow = c(1, 1))
CDFPlot(xFit)
dev.off()
}
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
message("\n")
}
if(fSeverity==TRUE){
f <- "severity"
message(paste0("Start Fitting Line:",l," Type:",t," ",f))
nclosed <- nrow(fitdata[fitdata[,"status"]=="CLOSED",])
nvalid <- nrow(fitdata[fitdata[,"claimLiability"]==TRUE & fitdata[,"status"]=="CLOSED",])
aggP0 <- 1-nvalid/nclosed
p0Set <- fitdata[fitdata[,"status"]=="CLOSED",]
p0Set[,"settlementYears"] <- ceiling(as.numeric(as.Date(p0Set[,"settlementDate"]) - as.Date(p0Set[,"occurrenceDate"]))/365)
p0Set <- p0Set[,colnames(p0Set) %in% c("settlementYears","claimLiability")]
#p0Set <- read.csv("C:/temp/CAS/pset.csv")
yclosed <- aggregate(claimLiability ~ settlementYears, data=p0Set, FUN="length")
if(nrow(p0Set[p0Set$claimLiability==FALSE,])>0){
yinvalid <- aggregate(claimLiability ~ settlementYears, data=p0Set[p0Set$claimLiability==FALSE,], FUN="length")
vp0 <- rep(0,max(yclosed$settlementYears))
for (i in c(1:length(vp0))) {
denominator <- yclosed[yclosed$settlementYears == i,]$claimLiability[1]
numerator <- yinvalid[yinvalid$settlementYears == i,]$claimLiability[1]
vp0[i] <- numerator/denominator
vp0[i] <- min(1,vp0[i])
}
vp0[is.na(vp0)] <- 0
} else {
vp0 <- rep(0,max(yclosed$settlementYears))
}
for (i in c(1:length(vp0))) {
fitsumfac[rfac,] <- c(l,t,"p0",i,vp0[i],0)
rfac <- rfac + 1
}
co@p0 <- new("DevFac",FacID=paste0(l,t,"p0"),FacModel= FALSE,meanList =vp0,volList =rep(0,length(vp0)))
if(co@sdata == "CLOSED") {
fitdata <- fitdata[fitdata[,"claimLiability"]==TRUE & fitdata[,"status"]=="CLOSED",]
} else {
fitdata <- fitdata[fitdata[,"claimLiability"]==TRUE,]
}
if (iTotalLoss == TRUE) {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","totalLoss")][,c("settlementDate","totalLoss")]
} else {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","incurredLoss","Deductible","Limit")][,c("settlementDate","incurredLoss","Deductible","Limit")]
}
if (iTotalLoss == TRUE) {
incurredlosses[,2] <- pmax(0.01,incurredlosses[,2])
}
if(TRUE){#objName(co@reportLag)!="Empirical"){
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
if (iTotalLoss == FALSE) {
xFit@iDL = TRUE
}
xFit <- setFitdata(xFit)
losses <- xFit@fitdata
bestBIC <- 1e10
isso <- 0
for (i in c(1:length(continuousDist))){
so <- tryCatch({
setTrialDistErr(xFit) <- new(continuousDist[i])
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",continuousDist[i]," distribution failed to fit"))
-1
}
)
if(so==1) {
isso <- 1
fitsummary[rfit,] <- c(l,t,f,xFit@soutput)
fitsummary[rfit,"p0"] <- aggP0
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
fitPlot(xFit)
dev.off()
}
} else {
fitsummary[rfit,] <- c(l,t,f,continuousDist[i],"mle",rep(NA,11))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,continuousDist[i],".jpg"))
par(mfrow = c(1, 1))
if (iTotalLoss == FALSE) {
plotText(paste0("Data cannot fit to truncated ",continuousDist[i]," distribution"))
} else {
plotText(paste0("Data cannot fit to ",continuousDist[i]," distribution"))
}
dev.off()
}
}
rfit <- rfit+1
if (i==1) {
bestfit <- xFit@fitted
if (!is.na(xFit@soutput[1,12])) {bestBIC <- as.numeric(xFit@soutput[1,12])}
} else if ((!is.na(xFit@soutput[1,12])) & (as.numeric(xFit@soutput[1,12])<bestBIC)){
bestfit <- xFit@fitted
bestBIC <- as.numeric(xFit@soutput[1,12])
}
}
if (isso == 1) {
co@severity <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@severity)))
xFit@fitted<-bestfit
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
} else {
co@severity@fitsucc <- FALSE
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No distribution is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
plotText(paste0("Data cannot fit to any tested distribution"))
dev.off()
}
} else {
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
co@severity <- xFit@fitted
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"))
fitPlot(xFit)
dev.off()
}
if (check == TRUE) {
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
xFit <- setFitdata(xFit)
setTrialDist(xFit) <- new("Empirical")
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,"Empirical.jpg"))
fitPlot(xFit)
dev.off()
cn <- paste0(l,t,f,"emp")
empiricals[,cn]<-Quantile(xFit@fitted,seq(0.001,1,0.001))
}
fitdata <- fitdata[fitdata[,"claimLiability"]==TRUE & fitdata[,"status"]=="CLOSED",]
reportlags <- as.numeric(as.Date(fitdata[,"reportDate"])-as.Date(fitdata[,"occurrenceDate"]))
reportlags <- ifelse(reportlags==0, runif(length(reportlags)),reportlags)
settlementlags <- as.numeric(as.Date(fitdata[,"settlementDate"])-as.Date(fitdata[,"reportDate"]))
settlementlags <- ifelse(settlementlags==0, runif(length(settlementlags)),settlementlags)
if (iTotalLoss == TRUE) {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","totalLoss")][,c("settlementDate","totalLoss")]
} else {
incurredlosses <- fitdata[,colnames(fitdata) %in% c("settlementDate","incurredLoss","Deductible","Limit")][,c("settlementDate","incurredLoss","Deductible","Limit")]
}
if (iTotalLoss == TRUE) {
incurredlosses[,2] <- pmax(0.01,incurredlosses[,2])
}
xFit <- new("FitDist", observation=incurredlosses, trend=co@severityIndex, startDate=startDate, method="mle",ifreq=FALSE)
if (iTotalLoss == FALSE) {
xFit@iDL = TRUE
}
xFit <- setFitdata(xFit)
losses <- xFit@fitdata
message("\n")
}
if(co@iCopula==TRUE & fReportLag==TRUE & fSettlementLag==TRUE & fSeverity==TRUE & fSSRCorrelation==TRUE){
f <- "ssrCorrelation"
message(paste0("Start Fitting Line:",l," Type:",t," ","Correlation among severity, report lag and settlement lag"))
x <- cbind(losses,settlementlags,reportlags)
if (nrow(x) > 2000) {
x <- x[sample(nrow(x), 2000), ]
}
if(nrow(x)>10) {
dist1 <- co@severity
dist2 <- co@settlementLag
dist3 <- co@reportLag
bestP <- 0
issrso <- 0
for (i in c(1:length(copulaList))){
if(copulaList[i]=="normal"){
nom.cop <- new("CopulaObj", type="normal",param=c(0.5,0.5,0.5),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,observation=x,fittest=copulaTest)
} else if(copulaList[i]=="t"){
nom.cop <- new("CopulaObj", type="t",param=c(0.5,0.5,0.5),df=5, marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,observation=x,fittest=copulaTest)
} else {
nom.cop <- new("CopulaObj", type=copulaList[i],param=c(3),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,observation=x,fittest=copulaTest)
}
so <- tryCatch({
if(nom.cop@type=="t") {nom.cop@fittest=TRUE} else {nom.cop@fittest=TRUE}#FALSE
nom.cop <- copulaFitErr(nom.cop)
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula failed to fit"))
-1
}
)
if(so==1) {
issrso <- 1
fitsumcop[rcop,] <- c(l,t,f,nom.cop@coutput)
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=240)
copulaFitPlot(nom.cop)
dev.off()
}
} else {
fitsumcop[rcop,] <- c(l,t,f,copulaList[i],"mpl",rep(NA,5))
nom.cop@coutput <- data.frame(matrix(c(copulaList[i],"mpl",rep(NA,5)),1,7))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=480)
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",copulaList[i]," copula"))
dev.off()
}
}
rcop <- rcop+1
if (i==1) {
bestfit <- nom.cop
if (!is.na(nom.cop@coutput[1,7])) {bestP <- as.numeric(nom.cop@coutput[1,7])}
} else if ((!is.na(nom.cop@coutput[1,7])) & (as.numeric(nom.cop@coutput[1,7])>bestP)){
bestfit <- nom.cop
bestP <- as.numeric(nom.cop@coutput[1,7])
}
}
if (issrso == 1) {
bestfit@fitsucc = TRUE
co@ssrCopula <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(co@ssrCopula)))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
copulaFitPlot(bestfit)
dev.off()
} else {
nom.cop <- new("CopulaObj", type="normal",param=c(0,0,0),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,fittest=copulaTest)
nom.cop@coutput = data.frame(Copula=character(),
Method=character(),
Parameter=character(),
SD=character(),
DoF=integer(),
Sn=double(),
p=double(),
stringsAsFactors=FALSE)
nom.cop@coutput[1,] = c("normal", "mpl", NA, NA, 3, NA, NA)
co@ssrCopula <- nom.cop
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No copula is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data cannot fit to any tested copula"))
dev.off()
}
message("\n")
} else {
dist1 <- co@severity
dist2 <- co@settlementLag
dist3 <- co@reportLag
nom.cop <- new("CopulaObj", type="normal",param=c(0,0,0),marginal=list(dist1=dist1,dist2=dist2,dist3=dist3),dimension=3,fittest=copulaTest)
nom.cop@coutput = data.frame(Copula=character(),
Method=character(),
Parameter=character(),
SD=character(),
DoF=integer(),
Sn=double(),
p=double(),
stringsAsFactors=FALSE)
nom.cop@coutput[1,] = c("normal", "mpl", NA, NA, 3, NA, NA)
co@ssrCopula <- nom.cop
message(paste0("Severity, settlement lag and report lag copula fitting is turned off due to insufficient data for line " , l, " type " , t))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data are insufficient fit to any tested copula"))
dev.off()
message("\n")
}
}
if(co@line==l & co@claimType==t) {object@claimobjs[[obji]]<-co}
} else {
message (paste0("Nothing to fit for line " , l, " type " , t))
message("\n")
}
}
}
if(fFreqCorrelation==TRUE & object@iCopula==TRUE){
tmp <- cbind(as.character(claimData$LoB),as.character(claimData$Type),claimData$occurrenceDate,as.character(substr(as.character(claimData$occurrenceDate),1,7)))
colnames(tmp) <- c("line","type","od","ym")
tmp <- aggregate(od ~ ym+line+type, data=tmp, FUN="length")
tmp <- cbind(tmp,as.numeric(substr(tmp$ym,1,4)),as.numeric(substr(tmp$ym,6,7)))
colnames(tmp) <- c("ym","line","type","od","year","mth")
for(l in object@lines){
for(t in object@types){
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if(co@line==l & co@claimType==t){
startYear <- as.numeric(substr(as.character(co@exposureIndex@startDate),1,4))
startMonth <- as.numeric(substr(as.character(co@exposureIndex@startDate),6,7))
endYear <- as.numeric(substr(as.character(evaluationDate),1,4))
endMonth <- as.numeric(substr(as.character(evaluationDate),6,7))
tmp$od <- ifelse((tmp$line==l & tmp$type==t),tmp$od/Probability(co@reportLag,(endYear-tmp$year)*365+(endMonth-tmp$mth)*30+15),tmp$od)
di <- (tmp$year - startYear)*12+(tmp$mth - startMonth)+1
di <- ifelse(di>360,360,ifelse(di<1,1,di))
tmp$od <- ifelse((tmp$line==l & tmp$type==t),tmp$od/co@exposureIndex@monthlyIndex[di],tmp$od)
tmp$od <- round(tmp$od)
}
}
}
tmp <- aggregate(od ~ ym+line, data=tmp, FUN="sum")
uym<-as.data.frame(as.character(unique(tmp$ym)))
colnames(uym)<-"uym"
for(l in object@lines){
uym[,l] <- 0
}
for(j in c(1:nrow(uym))){
for(l in object@lines){
uym[j,l] <- if(nrow(tmp[tmp$ym==uym$uym[j] & tmp$line==l,])==0) {0} else {tmp[tmp$ym==uym$uym[j] & tmp$line==l,]$od}
}
}
uym <- na.omit(uym[uym[2:ncol(uym)]>0,])
f <- "freqCorrelation"
l <- "Total"
t <- "Total"
message(paste0("Start Fitting frequency correlation among business lines"))
x <- data.matrix(uym[,-1])
if (nrow(x)>=10) {
bestP <- 0
ifcso <- 0
for (i in c(1:length(copulaList))){
if(copulaList[i]=="normal"){
nom.cop <- new("CopulaObj", type="normal",param=rep(0.5,(length(object@lines)-1)*length(object@lines)/2),dimension=length(object@lines),observation=x,fittest=copulaTest)
} else if(copulaList[i]=="t"){
nom.cop <- new("CopulaObj", type="t",param=rep(0.5,(length(object@lines)-1)*length(object@lines)/2),df=5, dimension=length(object@lines),observation=x,fittest=copulaTest)
} else {
nom.cop <- new("CopulaObj", type=copulaList[i],param=c(3),dimension=length(object@lines),observation=x,fittest=copulaTest)
}
so <- tryCatch({
nom.cop <- copulaFitErr(nom.cop)
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula fitted"))
1
}#,warning=function(w) {
# setTrialDist(xFit) <- new(continuousDist[i])
# return(1)
#}
,error=function(e) {
message(paste0("Line-", l, " Type-", t, " Fitting-", f,": ",copulaList[i]," copula failed to fit"))
-1
}
)
if(so==1) {
ifcso <- 1
fitsumcop[rcop,] <- c(l,t,f,nom.cop@coutput)
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=240)
copulaFitPlot(nom.cop)
dev.off()
}
} else {
fitsumcop[rcop,] <- c(l,t,f,copulaList[i],"mpl",rep(NA,5))
nom.cop@coutput <- data.frame(matrix(c(copulaList[i],"mpl",rep(NA,5)),1,7))
if (check == TRUE){
jpeg(filename = paste0(direc,"/fit/",object@fitfile,l,t,f,copulaList[i],".jpg"),width=480,height=480)
par(mfrow = c(1, 1))
plotText(paste0("Data cannot fit to ",copulaList[i]," copula"))
dev.off()
}
}
rcop <- rcop+1
if (i==1) {
bestfit <- nom.cop
if (!is.na(nom.cop@coutput[1,7])) {bestP <- as.numeric(nom.cop@coutput[1,7])}
} else if ((!is.na(nom.cop@coutput[1,7])) & (as.numeric(nom.cop@coutput[1,7])>bestP)){
bestfit <- nom.cop
bestP <- as.numeric(nom.cop@coutput[1,7])
}
}
if (ifcso == 1) {
object@freqCopula <- bestfit
message(paste0("Line:",l," Type:",t," ",f, " best fit: ",toString(object@freqCopula)))
png(filename = paste0(direc,"/fit/","fit",object@tag,f,".png"),width=480,height=240)
copulaFitPlot(bestfit)
dev.off()
} else {
nom.cop <- new("CopulaObj", type="normal",param=rep(0,(length(object@lines)-1)*length(object@lines)/2),dimension=length(object@lines),fittest=copulaTest)
nom.cop@coutput <- data.frame(matrix(c("normal","mpl",rep(NA,5)),1,7))
object@freqCopula <- nom.cop
message(paste0("Line:",l," Type:",t," ",f, " best fit: ","No copula is found appropriate."))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data cannot fit to any tested copula"))
dev.off()
}
} else {
nom.cop <- new("CopulaObj", type="normal",param=rep(0,(length(object@lines)-1)*length(object@lines)/2),dimension=length(object@lines),fittest=copulaTest)
nom.cop@coutput <- data.frame(matrix(c("normal","mpl",rep(NA,5)),1,7))
object@freqCopula <- nom.cop
message(paste0("Frequency copula fitting is turned off due to insufficient data"))
png(filename = paste0(direc,"/fit/","fit",object@tag,l,t,f,".png"),width=480,height=240)
plotText(paste0("Data is insufficient to fit to any tested copula"))
dev.off()
fFreqCorrelation = FALSE
}
}
if (check == TRUE) {
jpeg(filename = paste0(direc,"/fit/",object@fitfile,"na.jpg"),width=480,height=480)
par(mfrow = c(1, 1))
plotText(paste0("Not Fitted"))
dev.off()
jpeg(filename = paste0(direc,"/fit/",object@fitfile,"halfna.jpg"),width=480,height=240)
par(mfrow = c(1, 1))
plotText(paste0("Not Fitted"))
dev.off()
}
if(nrow(fitsummary)>=1) {write.table(fitsummary,paste0(direc,"/",fitname),row.names=FALSE, sep=",")}
if(ncol(fitsumfac)>=1) {write.table(fitsumfac,paste0(direc,"/",facname),row.names=FALSE, sep=",")}
if(nrow(fitsumcop)>=1) {write.table(fitsumcop,paste0(direc,"/",copname),row.names=FALSE, sep=",")}
if(ncol(empiricals)>=1) {write.table(empiricals,paste0(direc,"/",empname),row.names=FALSE, sep=",")}
if(object@iReport==TRUE){
#require(R2HTML)
pckgdir <- find.package("cascsim")
pckgdir <- paste0(pckgdir,"/css/")
if (!dir.exists(paste0(direc,"/fit"))) {dir.create(paste0(direc,"/fit"))}
copied <- file.copy(file.path(pckgdir, "report.css"), file.path(paste0(direc,"/fit/"),"report.css"))
if(object@fitRpt==""){
fitrpt <- paste0("sim",object@tag)
} else {
fitrpt <- paste0(object@fitRpt)
}
if (file.exists(paste0(direc,"/fit/",fitrpt,".html"))) {file.remove(paste0(direc,"/fit/",fitrpt,".html"))}
ReportBegin <- function(file, title, directory = direc) {
file = file.path(paste0(directory,"/fit"), file)
cat(paste("<html><head><title>", title, "</title></head>", "<body bgcolor=#D0D0D0>", sep = ""), file = file, append = FALSE)
}
ReportEnd <- function(file, directory = direc) {
file = file.path(paste0(directory,"/fit"), file)
cat("<hr size=1></body></html>", file = file, append = TRUE)
}
RptFetch<-function(line,type,dist){
if(dist=="ssrCorrelation" | dist=="freqCorrelation"){
result <- fitsumcop[fitsumcop$LoB==line & fitsumcop$Type==type & fitsumcop$Fit==dist,]
}else{
result <- fitsummary[fitsummary$LoB==line & fitsummary$Type==type & fitsummary$Fit==dist,]
}
rownames(result)<-NULL
return(result)
}
vecFlatten<-function(vec){
result<-""
if(length(vec)>1){
for(i in c(1:(length(vec)-1))){
result<-paste0(result,vec[i],", ")
}
result <- paste0(result,"and ",vec[length(vec)])
} else {
result <- vec[1]
}
return(result)
}
getCopulaInfo<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-object@lines
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
getCopulaSR<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-c("Severity","ReportLag","SettlementLag")
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
findClaimObj <- function(line,type,objpool){
for(co in objpool){
if(co@line==line & co@claimType==type) {
result <- co
break
} else {
result <- NULL
}
}
return(result)
}
HTMLcontent <- function(file, append = TRUE, directory = direc) {
file = file.path(paste0(directory,"/fit"), file)
cat("\n", file = file, append = append)
HTML.title("Claim Data Fitting Report", file=file, HR=1)
HTML(paste0("Claim Data Start Date: ",startDate,"; Claim Data Ending Date: ",evaluationDate), file=file)
HTML(paste0("Simulation Task ID: ",object@tag), file=file)
HTML(paste0("Report generated at ",date()), file=file)
HTMLhr(file = file)
HTML.title("Links",file = file, HR=4)
HTML(paste0("<a href=\"#tp\">Total Portfolio</a>"),file=file)
for(l in object@lines){
#HTML(paste0("<a href=\"#",l,"\">",l,"-Total","</a>"),file=file)
for(t in object@types){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
s<-paste0("<a href=\"#",l,t,"\">",l,"-",t,"</a>")
HTML(s,file=file)
}
}
}
HTMLhr(file = file)
HTML(paste0("<a name=\"tp\"></a>"), file = file)
HTML.title("Total Portfolio",file = file, HR=2)
HTML(paste0("The portfolio includes ",length(object@lines)," business line(s): ",vecFlatten(object@lines),"."),file=file)#,". Each business line may contain ",length(object@types)," type(s): ",vecFlatten(object@types)
if(fFreqCorrelation==TRUE & object@iCopula==TRUE){
HTML.title("Frequency Copula Among Business Lines",file = file, HR=3)
if (ifcso == 1) {
HTML(paste0("Best Copula: ",getCopulaInfo(object@freqCopula)$out), file = file)
if (!is.null(getCopulaInfo(object@freqCopula)$outm)){
HTML(getCopulaInfo(object@freqCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,"freqCorrelation.png"),Caption = "Copula Fitting Plot",file = file)
} else {
HTML(paste0("No tested copula is found appropriate."),file=file)
}
HTML("Tested Copulas", file = file)
HTML(RptFetch("Total","Total","freqCorrelation")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLhr(file = file)
for(l in object@lines){
HTML(paste0("<a name=\"",l,"\"></a>"), file = file)
for(t in object@types){
if (!is.null(findClaimObj(l,t,object@claimobjs))){
if(fFrequency==TRUE){
HTML(paste0("<a name=\"",l,t,"\"></a>"), file = file)
HTML.title(paste0(l,"-",t),file = file, HR=2)
HTML.title(paste0("Frequency"),file = file, HR=3)
#if (ifso == 1) {
if (findClaimObj(l,t,object@claimobjs)@frequency@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@frequency)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"frequency.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"frequency")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
}
if(fSeverity==TRUE){
HTML.title(paste0("Severity"),file = file, HR=3)
#if (isso == 1) {
if (findClaimObj(l,t,object@claimobjs)@severity@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"severity.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"severity")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Probability of Zero Payment by Development Year(p0): ",findClaimObj(l,t,object@claimobjs)@p0@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@p0), file = file, innerBorder=0.5, row.names=TRUE)
HTMLhr(file = file)
}
}
if(fReportLag==TRUE){
HTML.title(paste0("Report Lag"),file = file, HR=3)
#if (irlso == 1) {
if (findClaimObj(l,t,object@claimobjs)@reportLag@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reportLag)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"reportLag.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"reportLag")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
}
if(fSettlementLag==TRUE){
HTML.title(paste0("Settlement Lag"),file = file, HR=3)
#if (islso == 1) {
if (findClaimObj(l,t,object@claimobjs)@settlementLag@fitsucc == TRUE) {
HTML(paste0("Best Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@settlementLag)),file=file)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"settlementLag.png"),Caption = "",file = file)
HTML("Tested Distributions", file = file)
HTML(RptFetch(l,t,"settlementLag")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested distribution is found appropriate."),file=file)
}
HTMLhr(file = file)
}
if(fDeductible==TRUE){
HTML.title(paste0("Deductible Empirical Distribution"),file = file, HR=3)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"Deductible.png"),Caption = "",file = file,WidthHTML=250)
HTMLhr(file = file)
}
if(fLimit==TRUE){
HTML.title(paste0("Limit Empirical Distribution"),file = file, HR=3)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"Limit.png"),Caption = "",file = file,WidthHTML=250)
HTMLhr(file = file)
}
if(findClaimObj(l,t,object@claimobjs)@iCopula==TRUE & fReportLag==TRUE & fSettlementLag==TRUE & fSeverity==TRUE & fSSRCorrelation==TRUE){
HTML.title(paste0("Severity, Report Lag and Settlement Lag Copula"),file = file, HR=3)
#if (issrso == 1) {
if (findClaimObj(l,t,object@claimobjs)@ssrCopula@fitsucc == TRUE) {
HTML(paste0("Best Copula: ",getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$out), file = file)
HTML(getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
HTMLInsertGraph(paste0(direc,"/fit/","fit",object@tag,l,t,"ssrCorrelation.png"),Caption = "",file = file)
HTML("Tested Copulas", file = file)
HTML(RptFetch(l,t,"ssrCorrelation")[,-c(1:3)], file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML(paste0("No tested copula is found appropriate."),file=file)
}
HTMLhr(file = file)
}
}
}
}
message(paste("Report generated: ",file," at ",date(), "\n", sep = ""))
}
Report <- function(file) {
ReportBegin(file,"Claim Data Fitting Report")
if(copied == TRUE | file.exists(paste0(direc,"/fit/","report.css"))) {
HTMLCSS(file = paste0(direc,"/fit/",file), CSSfile = paste0(direc,"/fit/","report.css"))
}
HTMLcontent(file)
ReportEnd(file)
}
big_mark_opt <- getOption("R2HTML.format.big.mark")
decimal_mark_opt <- getOption("R2HTML.format.decimal.mark")
options("R2HTML.format.big.mark"=",")
options("R2HTML.format.decimal.mark"=".")
Report(paste0(fitrpt,".html"))
on.exit(options("R2HTML.format.big.mark"=big_mark_opt))
on.exit(options("R2HTML.format.decimal.mark"=decimal_mark_opt), add = TRUE)
}
} else {
stop("Reset iFit of the simulation object to TRUE to fit frequency, severity, report lag, settlement lag and frequency correlation to claim data.")
}
message("Claim data fitting done")
return(object)
} else {
message("Nothing to fit or no claim data.")
message("Claim data fitting done")
}
}, error = function(err){
message("Something is wrong. Please check the error messages")
if (exists("l") & exists("t") & exists("f")){
message(paste0(">>>Critical Error for claim distribution fitting: ",err," Line-", l, " Type-", t, " Fitting-", f))
} else {
message(paste0(">>>Critical Error for claim distribution fitting: ",err))
}
message("Claim data fitting done")
gc()
return(-1)
})
})
#' Claim simulation at line/type/status level
#' @name claimSimulation
#' @param object Simulation object
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' library(cascsim)
#' data(claimdata)
#' lines <- c("Auto")
#' types <- c("N")
#' AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' AutoN@frequency <- new("Poisson", p1 =50)
#' AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' AutoN@reportLag <- new("Exponential", p1 =0.1)
#' AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' AutoN@iCopula <- TRUE #use copula
#' AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' param = c(0.1,0.2,0.1))#A Gaussian Copula
#' AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' paras =c(5,1.5,0.005,1.2,3))
#' AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' simobj <- new("Simulation", lines=lines, types=types,
#' claimobjs= list(AutoN),workingFolder=tempdir())
#' simobj@simNo <- 1
#' simobj@iRBNER <-FALSE
#' simobj@iROPEN <-FALSE
#' simobj@iIBNR <-TRUE
#' simobj@iUPR <-FALSE
#' simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' @rdname claimSimulation-methods
#'
#' @importFrom utils write.table
#' @importFrom parallel clusterSetRNGStream makeCluster clusterExport parLapply stopCluster
#'
#' @exportMethod claimSimulation
setGeneric("claimSimulation", function(object, ...) standardGeneric("claimSimulation"))
#' @param claimData claim data including existing claims for RBNER and claim reopenness analysis;
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims).
#' @param append Boolean variable to indicate whether existing simulation results need to be kept.
#' @rdname claimSimulation-methods
#' @aliases claimSimulation,ANY-method
setMethod("claimSimulation", signature("Simulation"), function(object, claimData=data.frame(), startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"),append=TRUE)
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
#rn<-round(runif(1,1,10000))
if(object@simfile==""){
filename <- paste0("sim",object@tag,".csv")
} else {
filename <- paste0(object@simfile,".csv")
}
if(object@plog==""){
mcfile <- paste0("mcsim",object@tag,".txt")
} else {
mcfile <- paste0(object@plog,".txt")
}
simdata = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
iclass <- object@iRBNER | object@iROPEN | object@iIBNR | object@iUPR
if(length(object@lines)>0 & length(object@types)>0 & length(object@claimobjs)>0 & iclass) {
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if (file.exists(mcfile)) {file.remove(mcfile)}
if (append==FALSE & file.exists(filename)) {file.remove(filename)}
if(evaluationDate < startDate | futureDate < startDate){
stop("Evaluation date cannot be earlier than the start date.")
} else {
nmonths <- (as.numeric(substr(as.character(futureDate),1,4))-as.numeric(substr(as.character(startDate),1,4)))*12+as.numeric(substr(as.character(futureDate),6,7))-as.numeric(substr(as.character(startDate),6,7))+1
}
monthlyfreq <- function(mthfreq,startDate,endDate,objRptLag,iRptLag=FALSE) {
startyear<-as.numeric(substr(as.character(startDate),1,4))
startmonth<-as.numeric(substr(as.character(startDate),6,7))
endyear<-as.numeric(substr(as.character(endDate),1,4))
endmonth<-as.numeric(substr(as.character(endDate),6,7))
nmth<-(endyear-startyear)*12+endmonth-startmonth+1
istart <- max(1,(startyear-as.numeric(substr(as.character(startDate),1,4)))*12+startmonth-as.numeric(substr(as.character(startDate),6,7))+1)
iend <- min(istart + nmth - 1,nmonths)
result <- mthfreq[istart:iend]
if (iRptLag == FALSE) {
return(round(result))
} else {
for (i in c(1:(iend-istart+1))){
prob <- mean(Probability(objRptLag,(nmth-istart-i+1)*30+c(1:30)))
result[i] <- result[i]*(1-prob)
}
return(round(result))
}
}
if(object@ncores>1 & object@simNo >1){
#require(parallel)
pC <- makeCluster(object@ncores, outfile = mcfile)
vsim <- rep(0,object@ncores)
vsims <- rep(0,object@ncores)
vobj<-list()
if (object@ncores > object@simNo) {
object@ncores <- max(object@simNo,2)
}
for (j in c(1:object@ncores)){
vobj<- c(vobj,object)
vsim[j] <- floor(object@simNo/object@ncores)*j
vsims[j] <- floor(object@simNo/object@ncores)*(j-1)+1
}
vsim[object@ncores] <- object@simNo
#set.seed(123)
clusterSetRNGStream (cl = pC)
clusterExport(cl=pC, varlist=c("vsim", "vsims", "vobj", "claimData", "nmonths", "startDate", "evaluationDate", "futureDate", "monthlyfreq", "filename"), envir=environment())
msim <- function(i){
if(nchar(object@libpath)>0 & dir.exists(object@libpath)){
.libPaths(object@libpath)
library(cascsim,lib.loc=object@libpath)
}
simdata = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
if(object@simfile==""){
filenamei <- paste0("sim",object@tag,"c",i,".csv")
} else {
filenamei <- paste0(object@simfile,"c",i,".csv")
}
for(isim in c(vsims[i]:vsim[i])){
object<-vobj[[i]]
message(paste0("Worker ",i,",Simulation ",isim+object@startNo-1," started at ",date()))
if(object@iCopula==TRUE) {
freqs <- copulaSample(object@freqCopula,nmonths)
}
for(l in object@lines){
for(t in object@types){
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if(co@line==l & co@claimType==t){
co@simno<-as.character(isim+object@startNo-1)
if(object@iIBNR==TRUE){
if(object@iCopula==TRUE){
co@IBNRfreqIndex@startDate <- startDate
lid <- match(l,object@lines)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),startDate,evaluationDate,co@reportLag,TRUE))
} else {
co@IBNRfreqIndex@startDate <- startDate
mthfreq <- doSample(co@frequency,nmonths)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(mthfreq,startDate,evaluationDate,co@reportLag,TRUE))
}
}
if(object@iUPR==TRUE){
if(object@iCopula==TRUE){
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
lid <- match(l,object@lines)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
} else {
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
mthfreq <- doSample(co@frequency,nmonths)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(mthfreq,as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
}
}
stmp = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
co@iRBNER = object@iRBNER
co@iROPEN = object@iROPEN
co@iIBNR = object@iIBNR
co@iUPR = object@iUPR
stmp<-rbind(stmp,claimSample(co,claimData,startDate,evaluationDate)) #futureDate
if(file.exists(paste0(direc,"/",filenamei))){
write.table(stmp,paste0(direc,"/",filenamei),row.names=FALSE, sep=",",append = TRUE, col.names = FALSE)
}else{
write.table(stmp,paste0(direc,"/",filenamei),row.names=FALSE, sep=",",append = TRUE, col.names = TRUE)
}
simdata <- rbind(simdata,stmp)
}
}
}
# if (isim %% 10 == 0) {
message(paste0("Worker",i,",Simulation ",isim+object@startNo-1," ended at ",date()))
message(">>>Simulation is finished successfully<<<")
}
#file.rename("simdatatmp.csv", filename)
return(simdata)
}
#ptm<-proc.time()
#clusterExport(pC, "ptm")
msimdata <- parLapply(pC, 1:(vobj[[1]]@ncores), msim)
simdata = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
Paid=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
for (k in c(1:length(msimdata))){
simdata<-rbind(simdata,msimdata[[k]])
}
stopCluster (pC)
rm(msimdata)
gc()
if(file.exists(paste0(direc,"/",filename))){
write.table(simdata,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = FALSE)
}else{
write.table(simdata,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = TRUE)
}
for(i in c(1:object@ncores)){
if(object@simfile==""){
if(file.exists(paste0("sim",object@tag,"c",i,".csv"))) file.remove(paste0("sim",object@tag,"c",i,".csv"))
} else {
if(file.exists(paste0(object@simfile,"c",i,".csv"))) file.remove(paste0(object@simfile,"c",i,".csv"))
}
}
} else {
for(isim in c(1:object@simNo)){
if(object@iCopula==TRUE) {
freqs <- copulaSample(object@freqCopula,nmonths)
}
message(paste0("Simulation ",isim+object@startNo-1," started at ",date()))
for(l in object@lines){
for(t in object@types){
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if(co@line==l & co@claimType==t){
co@simno<-as.character(isim+object@startNo-1)
if(object@iIBNR==TRUE){
if(object@iCopula==TRUE){
co@IBNRfreqIndex@startDate <- startDate
lid <- match(l,object@lines)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),startDate,evaluationDate,co@reportLag,TRUE))
} else {
co@IBNRfreqIndex@startDate <- startDate
mthfreq <- doSample(co@frequency,nmonths)
co@IBNRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, startDate, evaluationDate)* monthlyfreq(mthfreq,startDate,evaluationDate,co@reportLag,TRUE))
}
}
if(object@iUPR==TRUE){
if(object@iCopula==TRUE){
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
lid <- match(l,object@lines)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(Quantile(co@frequency,freqs[,lid]),as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
} else {
co@UPRfreqIndex@startDate <- as.Date(evaluationDate)+15
mthfreq <- doSample(co@frequency,nmonths)
co@UPRfreqIndex@monthlyIndex <- round(shiftIndex(co@exposureIndex, as.Date(evaluationDate)+15, futureDate)* monthlyfreq(mthfreq,as.Date(evaluationDate)+15,futureDate,co@reportLag,FALSE))
}
}
stmp = data.frame(Sim=character(),
ClaimID=character(),
LoB=character(),
Type=character(),
status=character(),
occurrenceDate=character(),
reportDate=character(),
incurredLoss=double(),
osRatio=double(),
settlementDate=character(),
totalLoss=double(),
ultimateLoss=double(),
Deductible=double(),
Limit=double(),
Paid=double(),
reopenDate=character(),
resettleDate=character(),
reopenLoss=double(),
expectedLoss=double(),
LAE=double(),
ultimateLAE=double(),
expectedLAE=double(),
stringsAsFactors=FALSE)
co@iRBNER = object@iRBNER
co@iROPEN = object@iROPEN
co@iIBNR = object@iIBNR
co@iUPR = object@iUPR
stmp<-rbind(stmp,claimSample(co,claimData,startDate,evaluationDate)) #futureDate
if(file.exists(paste0(direc,"/",filename))){
write.table(stmp,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = FALSE)
}else{
write.table(stmp,paste0(direc,"/",filename),row.names=FALSE, sep=",",append = TRUE, col.names = TRUE)
}
simdata <- rbind(simdata,stmp)
# if (isim %% 10 == 0) {
# message(paste0("Simulation ",isim,": ",proc.time()[3]," elapsed"))
}
}
}
#write.csv(simdata,filename,row.names=FALSE)
message(paste0("Simulation ",isim+object@startNo-1," ended at ",date()))
message(">>>Simulation is finished successfully<<<")
}
# write.csv(simdata,filename,row.names=FALSE)
#file.rename("simdatatmp.csv", filename)
}
gc()
return(simdata)
} else {
stop("Nothing to simulate.")
}
}, error = function(err){
if (exists("l") & exists("t") & exists("c") & exists("isim")){
message(paste0(">>>Critical Error for claim simulation: ",err," Line-", l, " Type-", t, " Class-", c, " Sim No.-", isim+object@startNo-1))
} else {
message(paste0(">>>Critical Error for claim simulation: ",err))
}
message(">>>Simulation is finished with error")
gc()
return(-1)
})
})
#' Claim simulation result summary
#' @name simSummary
#' @param object Simulation object
#' @param simdata simulation data generated by claimSimulation
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' #run time is about 30s(>10s) and is commented out here to avoid long waiting time
#' #library(cascsim)
#' #data(claimdata)
#' #lines <- c("Auto")
#' #types <- c("N")
#' #AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' #AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' #AutoN@frequency <- new("Poisson", p1 =50)
#' #AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' #AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' #AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' #AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' #AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' #AutoN@reportLag <- new("Exponential", p1 =0.1)
#' #AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' #AutoN@iCopula <- TRUE #use copula
#' #AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' #param = c(0.1,0.2,0.1))#A Gaussian Copula
#' #AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' #AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' #paras =c(5,1.5,0.005,1.2,3))
#' #AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' #meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' #AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' #meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' #AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' #meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' #AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' #AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' #simobj <- new("Simulation", lines=lines, types=types,
#' #claimobjs= list(AutoN),workingFolder=tempdir())
#' #simobj@simNo <- 1
#' #simobj@iRBNER <-FALSE
#' #simobj@iROPEN <-FALSE
#' #simobj@iIBNR <-TRUE
#' #simobj@iUPR <-FALSE
#' #simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' #evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' #simSummary <- simSummary(simobj,simdata, startDate = as.Date("2012-01-01"))
#' @rdname simSummary-methods
#' @exportMethod simSummary
setGeneric("simSummary", function(object, simdata, ...) standardGeneric("simSummary"))
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims).
#' @rdname simSummary-methods
#' @aliases simSummary,ANY-method
setMethod("simSummary", signature("Simulation", "data.frame"), function(object, simdata, startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"))
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
if(object@sumfile==""){
summaryname <- paste0("sum",object@tag,".csv")
} else {
summaryname <- paste0(object@sumfile,".csv")
}
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if(nrow(simdata)==0) {stop("No simulation data to summarize")}
if (object@iSummary==TRUE){
message(paste0("Starting summarizing the simulation results: ",date()))
# if(length(object@lines)>1) {
idx1<-c(object@lines,"Total")
# } else {
# idx1<-object@lines
# }
# if(length(object@types)>1) {
idx2<-c(object@types,"Total")
# } else {
# idx2<-object@types
# }
# if(length(object@classes)>1) {
idx3 <- vector()
if (object@iRBNER==TRUE) {idx3 <- c(idx3,"IBNER")}
if (object@iROPEN==TRUE) {idx3 <- c(idx3,"ROPEN")}
if (object@iIBNR==TRUE) {idx3 <- c(idx3,"IBNR")}
if (object@iUPR==TRUE) {idx3 <- c(idx3,"UPR")}
idx3<-c(idx3,"Total")
# } else {
# idx3<-object@classes
# }
idx4<-c(as.numeric(substr(as.character(startDate),1,4)):as.numeric(substr(as.character(futureDate),1,4)))
idx4<-c(idx4,"Total")
idx5<-c("avg","sd","min",paste0("percentile ",c(10, 30, 50, 75, 80, 95, 99)),"max")
simSummary<-vector()
summaryNames <- c("LoB","Type","Class","AccidentYear","Measure","Frequency","AvgIncurredLoss","TotalIncurredLoss","AvgUltimateLoss","TotalUltimateLoss","TotalPaid","TotalLAE","PaidLAE","ReopenProb")
simdata$occurrenceYear <- as.numeric(substr(as.character(as.Date(simdata$occurrenceDate)),1,4))
transStatus<-function(class){
if(class=="IBNER"){
return("OPEN")
} else if (class=="ROPEN") {
return("CLOSED")
} else {
return(class)
}
}
for(i1 in idx1){
l<-i1
for(i2 in idx2){
t<-i2
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if((co@line==l & co@claimType==t) | (i1 == "Total") | (i2 == "Total")){
for(i3 in idx3){
c<-i3
#isim<--10000
for(i4 in idx4){
if(i1 != "Total"){
datatmp <- simdata[simdata[,"LoB"] == i1,]
if(i2 != "Total") {datatmp <- datatmp[datatmp[,"Type"] == i2,]}
if(i3 != "Total") {datatmp <- datatmp[datatmp[,"status"] == transStatus(i3),]}
if(i3 == "ROPEN") {
closeCount <- nrow(datatmp)/length(unique(datatmp$Sim))
datatmp <- datatmp[!is.na(datatmp$reopenLoss),]
}
# if(i3 == "Total") {
# datatmp <- datatmp[!(datatmp$status==transStatus("ROPEN") & is.na(datatmp$reopenLoss)),]
# }
if(i4 != "Total") {datatmp <- datatmp[datatmp[,"occurrenceYear"] == as.numeric(i4),]}
}else{
datatmp <- simdata
if(i2 != "Total") {datatmp <- datatmp[datatmp[,"Type"] == i2,]}
if(i3 != "Total") {datatmp <- datatmp[datatmp[,"status"] == transStatus(i3),]}
if(i3 == "ROPEN") {
closeCount <- nrow(datatmp)/length(unique(datatmp$Sim))
datatmp <- datatmp[!is.na(datatmp$reopenLoss),]
}
if(i4 != "Total") {datatmp <- datatmp[datatmp[,"occurrenceYear"] == as.numeric(i4),]}
#datatmp <- datatmp[!(datatmp$status==transStatus("ROPEN") & is.na(datatmp$reopenLoss)),]
}
if (nrow(datatmp)>0){
colreserve <- c("Sim","status","incurredLoss","osRatio","ultimateLoss","reopenLoss","LAE","ultimateLAE")
datatmp <- datatmp[,colreserve]
datatmp$paid <- datatmp$incurredLoss*(1-datatmp$osRatio)
datatmp$ultimateLoss <- ifelse(datatmp$status=="CLOSED" & !is.na(datatmp$reopenLoss),datatmp$reopenLoss,datatmp$ultimateLoss)
datatmp$ultimateLAE <- ifelse(datatmp$status=="CLOSED" & is.na(datatmp$reopenLoss),datatmp$LAE,datatmp$ultimateLAE)
datatmp$status<-NULL
datatmp$reopenLoss<-NULL
frequency<-aggregate(incurredLoss ~ Sim, data = datatmp, length)[,2]
avgIncurredLoss<-aggregate(incurredLoss ~ Sim, data = datatmp, mean)[,2]
totalIncurredLoss<-aggregate(incurredLoss ~ Sim, data = datatmp, sum)[,2]
avgUltimateLoss<-aggregate(ultimateLoss ~ Sim, data = datatmp, mean)[,2]
totalUltimateLoss<-aggregate(ultimateLoss ~ Sim, data = datatmp, sum)[,2]
totalPaidLoss<-aggregate(paid ~ Sim, data = datatmp, sum)[,2]
totalLAE<-aggregate(ultimateLAE ~ Sim, data = datatmp, sum)[,2]
totalPaidLAE<-aggregate(LAE ~ Sim, data = datatmp, sum)[,2]
if(i3=="ROPEN") reopenProb<-aggregate(incurredLoss ~ Sim, data = datatmp, length)[,2]/closeCount
gc()
for(i5 in idx5){
if(i5 == "avg"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,mean(frequency),mean(avgIncurredLoss),mean(totalIncurredLoss),mean(avgUltimateLoss),mean(totalUltimateLoss),mean(totalPaidLoss),mean(totalLAE),mean(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,mean(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else if(i5 == "sd"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,sd(frequency),sd(avgIncurredLoss),sd(totalIncurredLoss),sd(avgUltimateLoss),sd(totalUltimateLoss),sd(totalPaidLoss),sd(totalLAE),sd(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,sd(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else if(i5 == "min"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,min(frequency),min(avgIncurredLoss),min(totalIncurredLoss),min(avgUltimateLoss),min(totalUltimateLoss),min(totalPaidLoss),min(totalLAE),min(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,min(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else if(i5 == "max"){
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,max(frequency),max(avgIncurredLoss),max(totalIncurredLoss),max(avgUltimateLoss),max(totalUltimateLoss),max(totalPaidLoss),max(totalLAE),max(totalPaidLAE))
if(i3=="ROPEN"){
tl <- c(tl,max(reopenProb))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
} else {
pc <- as.numeric(substr(i5,12,nchar(i5)))/100
tl<- c(i1,i2,i3,i4,i5)
tl <- c(tl,quantile(frequency,pc),quantile(avgIncurredLoss,pc),quantile(totalIncurredLoss,pc),quantile(avgUltimateLoss,pc),quantile(totalUltimateLoss,pc),quantile(totalPaidLoss,pc),quantile(totalLAE,pc),quantile(totalPaidLAE,pc))
if(i3=="ROPEN"){
tl <- c(tl,quantile(reopenProb,pc))
} else {
tl <- c(tl,NA)
}
names(tl)<-summaryNames
}
simSummary<-rbind(simSummary,tl)
}
}
}
}
}
}
message(paste0("Business Line ",i1," summarized: ",date()))
}
rownames(simSummary)<-NULL
simSummary<-as.data.frame(simSummary)
#simSummary$AccidentYear=round(as.numeric(as.character(simSummary$AccidentYear)))
simSummary$Frequency=round(as.numeric(as.character(simSummary$Frequency)))
simSummary$AvgIncurredLoss=round(as.numeric(as.character(simSummary$AvgIncurredLoss)),2)
simSummary$TotalIncurredLoss=round(as.numeric(as.character(simSummary$TotalIncurredLoss)),2)
simSummary$AvgUltimateLoss=round(as.numeric(as.character(simSummary$AvgUltimateLoss)),2)
simSummary$TotalUltimateLoss=round(as.numeric(as.character(simSummary$TotalUltimateLoss)),2)
simSummary$TotalPaid=round(as.numeric(as.character(simSummary$TotalPaid)),2)
simSummary$TotalLAE=round(as.numeric(as.character(simSummary$TotalLAE)),2)
simSummary$PaidLAE=round(as.numeric(as.character(simSummary$PaidLAE)),2)
simSummary$ReopenProb <- ifelse(is.na(simSummary$ReopenProb),NA,round(as.numeric(as.character(simSummary$ReopenProb)),4))
write.table(simSummary,paste0(direc,"/",summaryname),row.names=FALSE, sep=",")
return(simSummary)
} else {
stop("Reset iSummary of the simulation object to TRUE to start summarizing the result.")
}
}, error = function(err){
if (exists("l") & exists("t") & exists("c")){
message(paste0(">>>Critical Error for claim simulation summarization: ",err," Line-", l, " Type-", t, " Class-", c))
} else {
message(paste0(">>>Critical Error for claim simulation summarization: ",err))
}
return(-1)
gc()
})
})
#' Claim simulation result triangles
#' @name simTriangle
#' @param object Simulation object
#' @param claimdata claim data used as basis for simulation
#' @param simdata simulation data generated by claimSimulation
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' #run time is about 30s(>10s) and is commented out here to avoid long waiting time
#' #library(cascsim)
#' #data(claimdata)
#' #lines <- c("Auto")
#' #types <- c("N")
#' #AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' #AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' #AutoN@frequency <- new("Poisson", p1 =50)
#' #AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' #AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' #AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' #AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' #AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' #AutoN@reportLag <- new("Exponential", p1 =0.1)
#' #AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' #AutoN@iCopula <- TRUE #use copula
#' #AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' #param = c(0.1,0.2,0.1))#A Gaussian Copula
#' #AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' #AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' #paras =c(5,1.5,0.005,1.2,3))
#' #AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' #meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' #AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' #meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' #AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' #meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' #AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' #AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' #simobj <- new("Simulation", lines=lines, types=types,
#' #claimobjs= list(AutoN),workingFolder=tempdir())
#' #simobj@simNo <- 1
#' #simobj@iRBNER <-FALSE
#' #simobj@iROPEN <-FALSE
#' #simobj@iIBNR <-TRUE
#' #simobj@iUPR <-FALSE
#' #simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' #evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' #simSummary <- simSummary(simobj,simdata, startDate = as.Date("2012-01-01"))
#' #simTriangle <- simTriangle(simobj,claimdata,simdata, startDate = as.Date("2012-01-01"))
#' @rdname simTriangle-methods
#' @importFrom methods new
#' @exportMethod simTriangle
setGeneric("simTriangle", function(object, claimdata, simdata, ...) standardGeneric("simTriangle"))
#' @param frequency triangle frequency, either "yearly" or "quarterly";
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims).
#' @rdname simTriangle-methods
#' @aliases simTriangle,ANY-method
setMethod("simTriangle", signature("Simulation", "data.frame", "data.frame"), function(object, claimdata, simdata, frequency="yearly", startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"))
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
if(object@sumfile==""){
summaryname <- paste0("tri",object@tag,".csv")
} else {
summaryname <- paste0("tri",object@sumfile,".csv")
}
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this Simulation object")
}
if (nrow(claimdata)>0){
claimdata <- claimdata[claimdata[,"LoB"] %in% object@lines & claimdata[,"Type"] %in% object@types,]
claimdata[,"status"] = toupper(claimdata[,"status"])
claimdata[,"occurrenceDate"] = toDate(claimdata[,"occurrenceDate"])
claimdata[,"reportDate"] = toDate(claimdata[,"reportDate"])
claimdata[,"settlementDate"] = toDate(claimdata[,"settlementDate"])
}
if (object@iSummary==TRUE){
message(paste0("Starting creating triangles from the simulation results: ",date()))
# if(length(object@lines)>1) {
idx1<-c(object@lines,"Total")
# } else {
# idx1<-object@lines
# }
# if(length(object@types)>1) {
idx2<-c(object@types,"Total")
# } else {
# idx2<-object@types
# }
idx3<-c("reportedCount","closedCount","incurredLoss")
# idx4<-c(as.numeric(substr(as.character(startDate),1,4)):as.numeric(substr(as.character(futureDate),1,4)))
idx5<-c("upper","avg",paste0("percentile ",c(10,50,90,99)))
if (object@simNo < 10) {
idx5<-c("upper","avg")
}
if (nrow(claimdata)==0 & object@simNo >=10) {
idx5<-c("avg",paste0("percentile ",c(10,50,90,99)))
}
if (nrow(claimdata)==0 & object@simNo <10) {
idx5<-c("avg")
}
#
xTri <- new("Triangle", triID = "TRI1", type = "closedCount", startDate=startDate, frequency="yearly", sim=0, iRBNER=object@iRBNER, iROPEN=object@iROPEN)
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,evaluationDate=evaluationDate,futureDate=futureDate)
dynames <- colnames(xTri@rectangle)
simSummary<-vector()
summaryNames <- c("LoB","Type","Triangle","Measure","AccidentPeriod",dynames)
for(i1 in idx1){
l<-i1
for(i2 in idx2){
t<-i2
for(co in object@claimobjs){
if(co@line==l & co@claimType==t) break
}
if((co@line==l & co@claimType==t) | (i1 == "Total") | (i2 == "Total")){
for(i3 in idx3){
c<-i3
xTri@type <- c
for(i5 in idx5){
if (i5 == "upper"){
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
cname <- colnames(xTri@upper)
rname <- rownames(xTri@upper)
tl<- data.frame(xTri@upper)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else if(i5 == "avg"){
xTri@sim <- 0
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else if(i5 == "min"){
xTri@sim <- NaN
xTri@percentile <- 0
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else if(i5 == "max"){
xTri@sim <- NaN
xTri@percentile <- 100
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
} else {
xTri@sim <- NaN
xTri@percentile <- as.numeric(substr(i5,12,nchar(i5)))
if (nrow(claimdata)>0) {
xTri<-setUpperTriangle(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
xTri<-setUpperKeep(xTri,claimdata,lob=i1,ctype=i2,evaluationDate=evaluationDate)
}
xTri<-setRectangle(xTri,simdata,lob=i1,ctype=i2,evaluationDate=evaluationDate,futureDate=futureDate)
cname <- colnames(xTri@rectangle)
rname <- rownames(xTri@rectangle)
tl<- data.frame(xTri@rectangle)
colnames(tl) <- cname
for (iname in dynames) {
if(!(iname %in% cname)){
tl[,iname]<-0
}
}
tl$AccidentPeriod <- rname
}
tl$LoB <- i1
tl$Type <- i2
tl$Triangle <- i3
tl$Measure <- i5
tl <- tl[,summaryNames]
simSummary<-rbind(simSummary,tl)
}
}
}
}
message(paste0("Business Line ",i1," triangle constructed: ",date()))
}
rownames(simSummary)<-NULL
write.table(simSummary,paste0(direc,"/",summaryname),row.names=FALSE, sep=",")
return(simSummary)
} else {
stop("Reset iSummary of the simulation object to TRUE to start summarizing the triangles.")
}
}, error = function(err){
if (exists("l") & exists("t") & exists("c") & exists("i5")){
message(paste0(">>>Critical Error for constructing claim simulation triangles: ",err," Line-", l, " Type-", t, " Class-", c, " Measure-",i5))
} else {
message(paste0(">>>Critical Error for constructing claim simulation triangles: ",err))
}
return(-1)
gc()
})
})
#' Generate claim simulation result report in html
#' @name simReport
#' @param object ClaimType object
#' @param simSummary simulation result summary generated by simSummary
#' @param ... Additional parameters that may or may not be used.
#' @examples
#' #run time is about 30s(>10s) and is commented out here to avoid long waiting time
#' #library(cascsim)
#' #data(claimdata)
#' #lines <- c("Auto")
#' #types <- c("N")
#' #AutoN <- new("ClaimType", line = "Auto", claimType = "N")
#' #AutoN@exposureIndex <- setIndex(new("Index",indexID="I1",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0)) # level exposure across time
#' #AutoN@frequency <- new("Poisson", p1 =50)
#' #AutoN@severityIndex <- setIndex(new("Index",indexID="I2",tabulate= FALSE,
#' #startDate=as.Date("2012-01-01"), annualizedRate = 0.02)) #assuming a 2% annual inflation
#' #AutoN@severity <- new("Lognormal", p1 =2, p2 =3)
#' #AutoN@deductible <- new("Empirical", empirical=matrix(c(0,1,100,100),2,2))
#' #AutoN@limit <- new("Empirical", empirical=matrix(c(0,1,1e8,1e8),2,2))
#' #AutoN@p0<-new("DevFac",meanList=c(0,0),volList=c(0,0))
#' #AutoN@reportLag <- new("Exponential", p1 =0.1)
#' #AutoN@settlementLag <- new("Exponential", p1 =0.05)
#' #AutoN@iCopula <- TRUE #use copula
#' #AutoN@ssrCopula <- new("CopulaObj", type ="normal", dimension = 3,
#' #param = c(0.1,0.2,0.1))#A Gaussian Copula
#' #AutoN@ssrCopula@marginal <- c(AutoN@severity,AutoN@settlementLag,AutoN@reportLag)
#' #AutoN@laeDevFac <- new("DevFac",FacID="F1",FacModel= TRUE,fun="linear",
#' #paras =c(5,1.5,0.005,1.2,3))
#' #AutoN@fIBNER <- new("DevFac",FacID="D1",FacModel= FALSE,
#' #meanList =c(1.2,1.15,1.1,1.05,1),volList =c(0,0,0,0,0))
#' #AutoN@reopen <- new("DevFac",FacID="D2",FacModel= FALSE,
#' #meanList =c(0.02,0.015,0.01,0.005,0),volList =c(0.003, 0.002, 0.001, 0.001, 0))
#' #AutoN@roDevFac <- new("DevFac",FacID="D3",FacModel= FALSE,
#' #meanList =c(1.05,1.1,1,1,1),volList =c(0.00589,0.0037,0.00632,0.00815,0))
#' #AutoN@reopenLag <- new("Exponential", p1 =0.01)
#' #AutoN@resettleLag <- new("Exponential", p1 =0.25)
#' #simobj <- new("Simulation", lines=lines, types=types,
#' #claimobjs= list(AutoN),workingFolder=tempdir())
#' #simobj@simNo <- 1
#' #simobj@iRBNER <-FALSE
#' #simobj@iROPEN <-FALSE
#' #simobj@iIBNR <-TRUE
#' #simobj@iUPR <-FALSE
#' #simdata <- claimSimulation(simobj,claimdata, startDate = as.Date("2012-01-01"),
#' #evaluationDate = as.Date("2016-12-31"), futureDate = as.Date("2017-12-31"))
#' #simSummary <- simSummary(simobj,simdata, startDate = as.Date("2012-01-01"))
#' #simTriangle <- simTriangle(simobj,claimdata,simdata, startDate = as.Date("2016-01-01"))
#' #simReport(simobj, simSummary, simTriangle, startDate = as.Date("2012-01-01"))
#' @rdname simReport-methods
#'
#' @importFrom R2HTML HTML.title HTML HTMLhr HTMLInsertGraph HTMLCSS
#' @import scatterplot3d
#'
#' @exportMethod simReport
setGeneric("simReport", function(object, simSummary, ...) standardGeneric("simReport"))
#' @param simTriangle triangle summary generated by simTriangle;
#' @param startDate Date after which claims are analyzed;
#' @param evaluationDate Date of evaluation for existing claims and IBNR;
#' @param futureDate Date of evaluation for UPR (future claims);
#' @param iYear Boolean that indicates whether summary by accident year should be produced in the report.
#' @rdname simReport-methods
#' @aliases simReport,ANY-method
setMethod("simReport", signature("Simulation", "data.frame"), function(object, simSummary, simTriangle=NA, startDate=as.Date("2012-01-01"),evaluationDate=as.Date("2016-12-31"),futureDate=as.Date("2017-12-31"),iYear=FALSE)
{
tryCatch({
startDate <- toDate(startDate)
evaluationDate <- toDate(evaluationDate)
futureDate <- toDate(futureDate)
if (object@iReport==TRUE){
if(nchar(object@workingFolder)>0 & dir.exists(object@workingFolder)){
direc <- object@workingFolder
} else if(nchar(object@workingFolder)>0 ){
dir.create(object@workingFolder)
direc <- object@workingFolder
} else {
stop("Please input a working folder for this simulation object")
}
classes <- vector()
if (object@iRBNER==TRUE) {classes <- c(classes,"IBNER")}
if (object@iROPEN==TRUE) {classes <- c(classes,"REOPEN")}
if (object@iIBNR==TRUE) {classes <- c(classes,"IBNR")}
if (object@iUPR==TRUE) {classes <- c(classes,"UPR")}
#require(R2HTML)
pckgdir <- find.package("cascsim")
pckgdir <- paste0(pckgdir,"/css/")
if (!dir.exists(paste0(direc,"/report"))) {dir.create(paste0(direc,"/report"))}
copied <- file.copy(file.path(pckgdir, "report.css"), file.path(paste0(direc,"/report/"),"report.css"))
if(object@htmlRpt==""){
htmlfile <- paste0("sim",object@tag)
} else {
htmlfile <- paste0(object@htmlRpt)
}
if (file.exists(paste0(direc,"/report/",htmlfile,".html"))) {file.remove(paste0(direc,"/report/",htmlfile,".html"))}
ReportBegin <- function(file, title, directory = direc) {
file = file.path(paste0(directory,"/report"), file)
cat(paste("<html><head><title>", title, "</title></head>", "<body bgcolor=#D0D0D0>", sep = ""), file = file, append = FALSE)
}
ReportEnd <- function(file, directory = direc) {
file = file.path(paste0(directory,"/report"), file)
cat("<hr size=1></body></html>", file = file, append = TRUE)
}
haveData<-function(line,type,class){
result <- simSummary[simSummary$LoB==line & simSummary$Type==type & simSummary$Class==class,]
if (nrow(result) > 0) {
result <- TRUE
} else {
result <- FALSE
}
return (result)
}
RptFetch<-function(line,type,class,accYear=TRUE){
result <- simSummary[simSummary$LoB==line & simSummary$Type==type & simSummary$Class==class,]
if (nrow(result) > 0) {
rowR<-c("avg","sd","min","percentile 10","percentile 30","percentile 50","percentile 75","percentile 90","percentile 95","percentile 99","max")#"percentile 99.5"
result <- subset(result, result$Measure %in% rowR)
if (accYear==FALSE) {result <- result[result$AccidentYear=="Total",]}
rownames(result)<-NULL
if(accYear==TRUE){
result <- result[,colnames(result) %in% c("AccidentYear","Measure","Frequency","TotalUltimateLoss")]
yrs <- unique(result$AccidentYear)
seps <- list()
coln <- vector()
for(y in c(1:length(yrs))){
tmp <- result[result$AccidentYear==yrs[y],]
if(y==1){
tmp <- tmp[,colnames(tmp) %in% c("Measure","Frequency","TotalUltimateLoss")]
coln <- c(coln,"Measure",paste0(yrs[y],"Count"),paste0(yrs[y],"Loss"))
seps<-c(seps,tmp)
} else {
tmp <- tmp[,colnames(tmp) %in% c("Frequency","TotalUltimateLoss")]
coln <- c(coln,paste0(yrs[y],"Count"),paste0(yrs[y],"Loss"))
seps<-c(seps,tmp)
}
}
for(y in c(1:length(seps))){
if(y==1){
result<-seps[[y]]
}else{
result<-cbind(result,seps[[y]])
}
}
result<-as.data.frame(result)
result[,1]<-rowR
colnames(result)<-coln
} else {
result <- result[,!colnames(result) %in% c("LoB","Type","Class","AccidentYear","AvgIncurredLoss","TotalIncurredLoss","TotalPaid","PaidLAE","ReopenProb")]
colnames(result) <- c("Measure","Count","AvgUltimateLoss","TotalUltimateLoss","TotalLAE")
}
} else {
result <- paste0("No ", class, " claims are simulated.")
}
return(result)
}
triangleFetch<-function(line,type,triangle,measure){
result <- simTriangle[simTriangle$LoB==line & simTriangle$Type==type & simTriangle$Triangle==triangle & simTriangle$Measure==measure,]
if (nrow(result) > 0) {
colremove <- c("LoB","Type","Triangle","Measure")
colkeep <- colnames(result)
colkeep <- colkeep[!colkeep %in% colremove]
if (measure == "upper") {colkeep <- colkeep[1:(nrow(result)+1)]}
result <- result[,colnames(result) %in% colkeep]
if (triangle=="incurredLoss" | triangle=="paidLoss"){
result[,!colnames(result) %in% c("AccidentPeriod")] <- round(result[,!colnames(result) %in% c("AccidentPeriod")]/1000)
}
for (i in c(2:ncol(result))){
result[,i] <- formatC(result[,i], format="d", big.mark=",")
}
if (measure == "upper") {
for (c in range(1:ncol(result))){
result[,c] = as.character(result[,c])
}
#result <- apply(result,2,as.character)
for(i in c(1:nrow(result))){
for(j in c(1:ncol(result))){
if(result[i,j]=="0"){result[i,j]=""}
}
}
} else {
result$AccidentPeriod <- as.character(result$AccidentPeriod)
}
colnames(result) <- colkeep
rownames(result) <- NULL
} else {
result <- NULL
}
return(result)
}
RptPlot<-function(line,type,class,accYear){
result <- simSummary[simSummary$LoB==line & simSummary$Type==type & simSummary$Class==class & simSummary$AccidentYear==accYear,]
##c("avg","sd","min",paste0("percentile ",c(10, 30, 50, 75, 80, 95, 99)),"max")
rowR<-paste0("percentile ",c(10, 30, 50, 75, 80, 95, 99))
result <- subset(result, result$Measure %in% rowR)
perc <- c(10, 30, 50, 75, 80, 95, 99)/100
par(mfrow = c(2, 2))
plot(result$TotalUltimateLoss, perc, type = "p", xlab = "TotalUltimateLoss", ylab = "Cumulative Probability", main = "Total Ultimate Loss CDF")
plot(result$AvgUltimateLoss, perc, type = "p", xlab = "AvgUltimateLoss", ylab = "Cumulative Probability", main = "Avg. Ultimate Loss CDF")
plot(result$Frequency, perc, type = "p", xlab = "Frequency", ylab = "Cumulative Probability", main = "Count CDF")
plot(result$TotalLAE, perc, type = "p", xlab = "TotalLAE", ylab = "Cumulative Probability", main = "Total LAE CDF")
}
distPlot<-function(claimobj){
par(mfrow = c(2, 2))
x<-doSample(claimobj@frequency, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@frequency, p)
plot(q, Density(claimobj@frequency, q),type="l", main = "Frequency", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@severity, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@severity, p)
plot(q, Density(claimobj@severity, q),type="l", main = "Severity", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@reportLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@reportLag, p)
plot(q, Density(claimobj@reportLag, q),type="l", main = "Report Lag", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@settlementLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@settlementLag, p)
plot(q, Density(claimobj@settlementLag, q),type="l", main = "Settlement Lag", xlab="q", ylab="density", col = "red")
}
idxPlot<-function(claimobj){
par(mfrow = c(2, 2))
plot(claimobj@exposureIndex@monthlyIndex, main = "Exposure Index", xlab="Time", ylab="Index")
plot(claimobj@severityIndex@monthlyIndex, main = "Severity Index", xlab="Time", ylab="Index")
if(claimobj@iCopula==TRUE){
samples<-copulaSample(claimobj@ssrCopula, 1000)
if (claimobj@ssrCopula@dimension==2)
{
plot(samples, col="blue")
}
else if (claimobj@ssrCopula@dimension>=2)
{
#require(scatterplot3d)
scatterplot3d(samples[,1:3], color="blue")
}
} else {
plotText("Severity and lags are independent.")
}
if (claimobj@laeDevFac@FacModel==TRUE){
plotText("LAE development factor is formula based.")
} else {
plot(claimobj@laeDevFac@meanList, main = "LAE Development", xlab="Duration", ylab="Factor")
}
}
roPlot<-function(claimobj){
par(mfrow = c(2, 2))
if (claimobj@reopen@FacModel==TRUE){
plotText("Reopen probability is formula based.")
} else {
plot(claimobj@reopen@meanList, main = "Reopen probability", xlab="Duration", ylab="Probability")
}
if (claimobj@roDevFac@FacModel==TRUE){
plotText("Reopen loss development is formula based.")
} else if (claimobj@irDevFac==3 & claimobj@roDevFac@distType == "normal"){
plot(claimobj@roDevFac@meanList, main = "Reopen loss development", xlab="Duration", ylab="Factor")
} else if (claimobj@irDevFac==3 & claimobj@roDevFac@distType != "normal"){
plotText(paste0("LDF based on ", claimobj@roDevFac@distType, " dist."))
} else {
plotText("Loss based on truncated severity dist.")
}
x<-doSample(claimobj@reopenLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@reopenLag, p)
plot(q, Density(claimobj@reopenLag, q),type="l", main = "Reopen Lag", xlab="q", ylab="density", col = "red")
x<-doSample(claimobj@resettleLag, 1000)
p<-seq(0,1, by=0.005)
q<-Quantile(claimobj@resettleLag, p)
plot(q, Density(claimobj@resettleLag, q),type="l", main = "Resettlement Lag", xlab="q", ylab="density", col = "red")
}
dlPlot<-function(claimobj){
par(mfrow = c(1, 2))
y<-claimobj@deductible
ysample<-doSample(y, 1000)
d1<-ecdf(ysample)
plot(d1, main = "CDF: Deductible", col="blue")
y<-claimobj@limit
ysample<-doSample(y, 1000)
d1<-ecdf(ysample)
plot(d1, main = "CDF: Limit", col="blue")
}
fibnerPlot<-function(claimobj){
par(mfrow = c(1, 1))
if (claimobj@fIBNER@FacModel==TRUE){
plotText("IBNER loss development is formula based.")
} else {
plot(claimobj@fIBNER@meanList, main = "RBNER Loss Development", xlab="Duration", ylab="Factor")
}
}
p0Plot<-function(claimobj){
par(mfrow = c(1, 1))
plot(claimobj@p0@meanList, main = "Probability of Zero Payment", xlab="Duration", ylab="Factor")
}
vecFlatten<-function(vec){
result<-""
if(length(vec)>1){
for(i in c(1:(length(vec)-1))){
result<-paste0(result,vec[i],", ")
}
result <- paste0(result,"and ",vec[length(vec)])
} else {
result <- vec[1]
}
return(result)
}
getCopulaInfo<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-object@lines
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
getCopulaSR<-function(c){
out <- paste0(c@info," ",c@type," copula, dimension=",c@dimension)
if(c@type=="t"){out<-paste0(out,", df=",c@df)}
if((c@type=="t" | c@type=="normal") & length(c@param)==(c@dimension*(c@dimension-1)/2)){
outm<-matrix(1,c@dimension,c@dimension)
irow<-2
icol<-1
for(cor in c@param){
outm[irow,icol]<-cor
outm[icol,irow]<-cor
if(irow<c@dimension){
irow<-irow+1
}else{
icol<-icol+1
irow<-icol+1
}
}
outm<-as.data.frame(outm)
colnames(outm)<-c("Severity","SettlementLag","ReportLag")
return(list(out=out,outm=outm))
}else{
out<-paste0(out,", parameter=",vecFlatten(c@param))
return(list(out=out))
}
}
findClaimObj <- function(line,type,objpool){
for(co in objpool){
if(co@line==line & co@claimType==type) {
result <- co
break
} else {
result <- NULL
}
}
return(result)
}
HTMLcontent <- function(file, append = TRUE, directory = direc) {
file = file.path(paste0(directory,"/report"), file)
cat("\n", file = file, append = append)
HTML.title("Claim Simulation Report", file=file, HR=1)
HTML(paste0("Claim Data Start Date: ",startDate,"; Claim Data Ending Date: ",evaluationDate,"; Future Claim Ending Date: ",futureDate), file=file)
HTML(paste0("Simulation Task ID: ",object@tag), file=file)
HTML(paste0("Report generated at ",date()), file=file)
HTMLhr(file = file)
HTML.title("Notation",file = file, HR=4)
HTML(paste0("AccidentPeriod: Accident Year or Quarter."),file=file)
HTML(paste0("Mxx: Development period. For example, M12 is the first development year and M24 is the second development year."),file=file)
HTML(paste0("Count: Number of simulated claims."),file=file)
HTML(paste0("AvgUltimateLoss: Ultimate Loss per claim. It is the average loss of all claims in one portfolio simulation. Each simulation has an average ultimate loss and multiple simulations allow the construction of the distribution."),file=file)
HTML(paste0("TotalUltimateLoss: Total Ultimate Loss for all simulated claims in one portfolio simulation. Multiple simulations allow the construction of its distribution."),file=file)
HTML(paste0("TotalLAE: Total Loss Adjustment Expense for all simulated claims in one portfolio simulation(if modeled separately from indemnity). Multiple simulations allow the construction of its distribution."),file=file)
HTML(paste0("Measure: Statistical measure such as avg(average), sd(standard deviation), min, max and a certain percentile."),file=file)
HTMLhr(file = file)
HTML.title("Links",file = file, HR=4)
HTML(paste0("<a href=\"#tp\">Total Portfolio</a>"),file=file)
for(l in object@lines){
HTML(paste0("<a href=\"#",l,"\">",l,"-Total","</a>"),file=file)
for(t in object@types){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
s<-paste0("<a href=\"#",l,t,"\">",l,"-",t,"</a>")
for(c in classes){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
s<-paste0(s," ","<a href=\"#",l,t,c,"\">",c,"</a>")
}
}
HTML(s,file=file)
}
}
}
HTMLhr(file = file)
HTML(paste0("<a name=\"tp\"></a>"), file = file)
HTML.title("Total Portfolio",file = file, HR=2)
HTML(paste0("The portfolio includes ",length(object@lines)," business line(s): ",vecFlatten(object@lines),". Each business line contains ",length(object@types)," type(s): ",vecFlatten(object@types),". ",length(classes)," class(es) of claims are simulated: ",vecFlatten(classes),"."),file=file)
if(!(is.na(simTriangle)[1])){
if (is.null(triangleFetch("Total","Total","reportedCount","upper"))) {
HTML.title("Average Simulated Reported Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","reportedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Average Simulated Closed Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","closedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Average Simulated Data Incurred Loss Rectangle (Thousands)",file = file, HR=3)
HTML(triangleFetch("Total","Total","incurredLoss","avg"), file = file, innerBorder=0.5, row.names=FALSE)
} else {
HTML.title("Claim Data Reported Count Triangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","reportedCount","upper"), file = file, innerBorder=0.5, row.names=FALSE)
HTML.title("Average Simulated Reported Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","reportedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Claim Data Closed Count Triangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","closedCount","upper"), file = file, innerBorder=0.5, row.names=FALSE)
HTML.title("Average Simulated Closed Count Rectangle",file = file, HR=3)
HTML(triangleFetch("Total","Total","closedCount","avg"), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("Claim Data Incurred Loss Triangle (Thousands)",file = file, HR=3)
HTML(triangleFetch("Total","Total","incurredLoss","upper"), file = file, innerBorder=0.5, row.names=FALSE)
HTML.title("Average Simulated Data Incurred Loss Rectangle (Thousands)",file = file, HR=3)
HTML(triangleFetch("Total","Total","incurredLoss","avg"), file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLhr(file = file)
}
HTML.title("Key Statistics - Total Portfolio, All Accident Years",file = file, HR=3)
HTML(RptFetch("Total","Total","Total",FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title("CDF - Total Portfolio, All Accident Years",file = file, HR=3)
png(filename = paste0(direc,"/report/","sim",object@tag,"tp.png"))
RptPlot("Total","Total","Total","Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,"tp.png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title("Key Statistics - Total Portfolio, By Accident Year",file = file, HR=3)
HTML(RptFetch("Total","Total","Total",TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
HTML.title("Annual Frequency Correlation",file = file, HR=3)
if(object@iCopula==FALSE){
HTML("Frequencies are assumed to be independent for different lines/types", file = file)
} else {
HTML(paste0("Frequencies are assumed to follow ",getCopulaInfo(object@freqCopula)$out), file = file)
HTML(getCopulaInfo(object@freqCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
HTMLhr(file = file)
for(l in object@lines){
HTML(paste0("<a name=\"",l,"\"></a>"), file = file)
HTML.title(l,file = file, HR=2)
HTML(RptFetch(l,"Total","Total",FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title(paste0("CDF - ",l,", All Accident Years"),file = file, HR=3)
png(filename = paste0(direc,"/report/","sim",object@tag,l,".png"))
RptPlot(l,"Total","Total","Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,".png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title(paste0("Key Statistics - ",l,", By Accident Year"),file = file, HR=3)
HTML(RptFetch(l,"Total","Total",TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
for(t in object@types){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("<a name=\"",l,t,"\"></a>"), file = file)
HTML.title(paste0(l,"-",t),file = file, HR=2)
HTML(RptFetch(l,t,"Total",FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
HTML.title(paste0("CDF - ",l,"-",t,", All Accident Years"),file = file, HR=3)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,".png"))
RptPlot(l,t,"Total","Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,".png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title(paste0("Key Statistics - ",l,"-",t,", By Accident Year"),file = file, HR=3)
HTML(RptFetch(l,t,"Total",TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Probability of Zero Payment by Development Year(p0): ",findClaimObj(l,t,object@claimobjs)@p0@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@p0), file = file, innerBorder=0.5, row.names=TRUE)
HTMLhr(file = file)
}
for(c in classes){
HTML(paste0("<a name=\"",l,t,c,"\"></a>"), file = file)
HTML.title(paste0(l,"-",t,"-",c),file = file, HR=3)
if (c=="IBNER"){
c2 = "IBNER"
} else if (c=="REOPEN") {
c2 = "ROPEN"
} else {
c2 = c
}
HTML(RptFetch(l,t,c2,FALSE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
if (haveData(l,t,c2) == TRUE) {
HTML.title(paste0("CDF - ",l,"-",t,"-",c,", All Accident Years"),file = file, HR=5)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,".png"))
RptPlot(l,t,c2,"Total")
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,".png"),Caption = "CDF",file = file)
HTMLhr(file = file)
if (iYear==TRUE) {
HTML.title(paste0("Key Statistics - ",l,"-",t,"-",c,", By Accident Year"),file = file, HR=5)
HTML(RptFetch(l,t,c,TRUE), file = file, innerBorder=0.5, row.names=FALSE)
HTMLhr(file = file)
}
HTML.title("Assumptions",file = file, HR=5)
if(c=="IBNER"){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
if(findClaimObj(l,t,object@claimobjs)@fIBNER@FacModel==TRUE & findClaimObj(l,t,object@claimobjs)@ioDevFac==3){
HTML(paste0("IBNER Loss Development: ",findClaimObj(l,t,object@claimobjs)@fIBNER@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@fIBNER)),file=file)
} else if (findClaimObj(l,t,object@claimobjs)@ioDevFac==3){
HTML(paste0("IBNER Loss Development: ",findClaimObj(l,t,object@claimobjs)@fIBNER@FacID, ". Distribution type: ", findClaimObj(l,t,object@claimobjs)@fIBNER@distType),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@fIBNER), file = file, innerBorder=0.5, row.names=TRUE)
} else if (findClaimObj(l,t,object@claimobjs)@ioDevFac==2){
HTML(paste0("IBNER Loss Development: Severity distribution floored by incurred loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
} else {
HTML(paste0("IBNER Loss Development: Severity distribution conditioned on paid loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
}
if (findClaimObj(l,t,object@claimobjs)@ioDevFac==3 & findClaimObj(l,t,object@claimobjs)@fIBNER@distType == "normal"){
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"fibner.png"),width=240,height=240)
fibnerPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"fibner.png"),Caption = "",file = file,WidthHTML=250)
}
}
}else if (c=="REOPEN"){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
if(findClaimObj(l,t,object@claimobjs)@reopen@FacModel==TRUE){
HTML(paste0("Reopen Probability: ",findClaimObj(l,t,object@claimobjs)@reopen@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@reopen)),file=file)
} else {
HTML(paste0("Reopen Probability: ",findClaimObj(l,t,object@claimobjs)@reopen@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@reopen), file = file, innerBorder=0.5, row.names=TRUE)
}
if(findClaimObj(l,t,object@claimobjs)@roDevFac@FacModel==TRUE & findClaimObj(l,t,object@claimobjs)@irDevFac==3){
HTML(paste0("Reopen Loss Development: ",findClaimObj(l,t,object@claimobjs)@roDevFac@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@roDevFac)),file=file)
} else if (findClaimObj(l,t,object@claimobjs)@irDevFac==3){
HTML(paste0("Reopen Loss Development: ",findClaimObj(l,t,object@claimobjs)@roDevFac@FacID, ". Distribution type: ", findClaimObj(l,t,object@claimobjs)@roDevFac@distType),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@roDevFac), file = file, innerBorder=0.5, row.names=TRUE)
} else if (findClaimObj(l,t,object@claimobjs)@irDevFac==2){
HTML(paste0("Reopen Loss Development: Severity distribution floored by incurred loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
} else {
HTML(paste0("Reopen Loss Development: Severity distribution conditioned on paid loss is used. ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
}
HTML(paste0("Reopen Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reopenLag)),file=file)
HTML(paste0("Resettlement Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@resettleLag)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"ro.png"))
roPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"ro.png"),Caption = "",file = file)
}
}else if (c=="IBNR"){
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Frequency Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@frequency)),file=file)
HTML(paste0("Severity Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
HTML(paste0("Report Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reportLag)),file=file)
HTML(paste0("Settlement Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@settlementLag)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"))
distPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"),Caption = "",file = file)
HTML(paste0("Severity Index: ",toString(findClaimObj(l,t,object@claimobjs)@severityIndex)),file=file)
HTML(paste0("Exposure Index: ",toString(findClaimObj(l,t,object@claimobjs)@exposureIndex)),file=file)
HTML(paste0("Correlation between Severity, settlement Lag and Report Lag: "),file=file)
if(findClaimObj(l,t,object@claimobjs)@iCopula==FALSE){
HTML("Severity, Settlement Lag and Report Lag are assumed to be independent.", file = file)
} else {
HTML(paste0("Severity, Settlement Lag and Report Lag are assumed to follow ",getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$out), file = file)
HTML(getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
if(findClaimObj(l,t,object@claimobjs)@laeDevFac@FacModel==TRUE){
HTML(paste0("LAE: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@laeDevFac)),file=file)
} else {
HTML(paste0("LAE Development Schedule: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@laeDevFac), file = file, innerBorder=0.5, row.names=TRUE)
}
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"))
idxPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"),Caption = "",file = file)
HTML(paste0("Deductible Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@deductible)),file=file)
HTML(paste0("Limit Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@limit)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"dl.png"),width=480,height=240)
dlPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"dl.png"),Caption = "",file = file)
}
}else if (c=="UPR" & haveData(l,t,"IBNR")==FALSE){ #sum("IBNR"==classes)==0 &
if(!is.null(findClaimObj(l,t,object@claimobjs))){
HTML(paste0("Frequency Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@frequency)),file=file)
HTML(paste0("Severity Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@severity)),file=file)
HTML(paste0("Report Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@reportLag)),file=file)
HTML(paste0("Settlement Lag Distribution: ",toString(findClaimObj(l,t,object@claimobjs)@settlementLag)),file=file)
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"))
distPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"dist.png"),Caption = "",file = file)
HTML(paste0("Severity Index: ",toString(findClaimObj(l,t,object@claimobjs)@severityIndex)),file=file)
HTML(paste0("Exposure Index: ",toString(findClaimObj(l,t,object@claimobjs)@exposureIndex)),file=file)
HTML(paste0("Correlation between Severity, settlement Lag and Report Lag: "),file=file)
if(findClaimObj(l,t,object@claimobjs)@iCopula==FALSE){
HTML("Severity, settlement Lag and Report Lag are assumed to be independent.", file = file)
} else {
HTML(paste0("Severity, settlement Lag and Report Lag are assumed to follow ",getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$out), file = file)
HTML(getCopulaSR(findClaimObj(l,t,object@claimobjs)@ssrCopula)$outm, file = file, innerBorder=0.5, row.names=FALSE)
}
if(findClaimObj(l,t,object@claimobjs)@laeDevFac@FacModel==TRUE){
HTML(paste0("LAE: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID," ",toString(findClaimObj(l,t,object@claimobjs)@laeDevFac)),file=file)
} else {
HTML(paste0("LAE Development Schedule: ",findClaimObj(l,t,object@claimobjs)@laeDevFac@FacID),file=file)
HTML(toString(findClaimObj(l,t,object@claimobjs)@laeDevFac), file = file, innerBorder=0.5, row.names=TRUE)
}
png(filename = paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"))
idxPlot(findClaimObj(l,t,object@claimobjs))
dev.off()
HTMLInsertGraph(paste0(direc,"/report/","sim",object@tag,l,t,c,"idx.png"),Caption = "",file = file)
}
} else {
HTML(paste0("Same as assumptions used for IBNR."),file=file)
}
}
HTMLhr(file = file)
}
}
}
}
message(paste("Report generated: ",file," at ",date(), "\n",sep = ""))
}
Report <- function(file) {
ReportBegin(file,"Claim Simulation Report")
if(copied == TRUE | file.exists(paste0(direc,"/report/","report.css"))) {
HTMLCSS(file = paste0(direc,"/report/",file), CSSfile = paste0(direc,"/report/","report.css"))
}
HTMLcontent(file)
ReportEnd(file)
}
big_mark_opt <- getOption("R2HTML.format.big.mark")
decimal_mark_opt <- getOption("R2HTML.format.decimal.mark")
options("R2HTML.format.big.mark"=",")
options("R2HTML.format.decimal.mark"=".")
Report(paste0(htmlfile,".html"))
on.exit(options("R2HTML.format.big.mark"=big_mark_opt))
on.exit(options("R2HTML.format.decimal.mark"=decimal_mark_opt), add = TRUE)
} else {
stop("Set iReport of simulation object to TRUE to generate html report.")
}
}, error = function(err){
message(paste0(">>>Critical Error for generating claim simulation report: ",err))
gc()
})
})
Try the cascsim package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.