R/dm_functions.R
In learnfin/learnfin: Statistical Learning Methods for Financial Decision Making
Defines functions
kmeans_learn
plot_export_kmeans
manual_learn
plot_export_manual
dm_learn
Documented in
dm_learn
kmeans_learn
manual_learn
plot_export_kmeans
plot_export_manual
#Below functions are individual DM algorithms and other supplementary functions used in uslfin package
#' @title K-Means Learn
#'
#' @description
#' Implements a specialized version of K-Means algorithm on the data set. When creating the clusters (in-sample) the function uses pricing_error, moneyness and maturity. But when predicting, it uses only moneyness and maturity covariates. All covariates are scaled between 0-100.
#'
#' @param raw_data The option data set given in the format of \link{uslfin_ds_1}.
#' @param CallPut It denotes whether to use the call or put options.
#' @param randseed To set the randomness seed to a known value. Good for reproducibility.
#' @param moneyness_interval Minimum and maximum of moneyness values. Required for rescaling.
#' @param maturity_interval Minimum and maximum of maturity values. Required for rescaling. In trading days (one year \= 252 days)
#' @param n_cluster Number of clusters. If \code{0}, then number of clusers is determined by the contract set size. Exact calculation is \code{min(round(nrow(training_matrix)/100),200)}
#' @export
kmeans_learn<-function(raw_data,CallPut="call",randseed=0,moneyness_interval=c(0.9,1.1),maturity_interval=c(4,252),n_cluster=0,export_plots=FALSE){
scaled_data <-
raw_data %>%
#Use only appropriate option type (call or put)
dplyr::filter(type == CallPut) %>%
#Rescale moneyness and maturity to a 0-100 scale each.
dplyr::mutate_(scaled_moneyness = lazyeval::interp(~((moneyness-min(moneyness_interval))/(max(moneyness_interval)-min(moneyness_interval))*100)),
scaled_maturity = lazyeval::interp(~((maturity-min(maturity_interval))/(max(maturity_interval)-min(maturity_interval))*100)))
#Prepare the training data
training_data <-
scaled_data %>%
#Filter only training data
dplyr::filter(t_or_p == "Training") %>%
dplyr::mutate_(scaled_error=lazyeval::interp(~(rescale_error(pricing_error,mederror=NA)*100)))
#Create the error + moneyness + maturity matrix for kmeans
training_matrix<-
training_data %>%
dplyr::select(scaled_error,scaled_moneyness,scaled_maturity) %>%
as.matrix
#Number of clusters are determined by the number of contracts (n_contracts/100). Max 200 clusters
if(n_cluster == 0){
n_cluster <- min(round(nrow(training_matrix)/100),200)
}
#Set the seed if there is a predetermined one
if(randseed>0){
set.seed(randseed)
}
#Train K-Means
kmeans_training <- kmeans(training_matrix,centers=n_cluster,nstart=round(n_cluster/5),iter.max=10^6)
#Get the cluster centers
kmeans_centers <- kmeans_training$centers[,c("scaled_moneyness","scaled_maturity")]
#Get the average pricing error of each cluster
kmeans_center_errors <-
training_data %>%
dplyr::mutate(clusters=kmeans_training$cluster) %>%
dplyr::group_by(clusters) %>%
dplyr::summarise_(model_estimate=lazyeval::interp(~mean(pricing_error,na.rm=TRUE)))
#Get both training and prediction contracts' prediction error on pricing errors.
result_data <-
scaled_data %>%
dplyr::rowwise() %>%
dplyr::mutate(clusters=get_the_cluster(scaled_moneyness,scaled_maturity,kmeans_centers)) %>%
dplyr::ungroup() %>%
dplyr::left_join(.,kmeans_center_errors,by="clusters") %>%
dplyr::mutate(prediction_error=abs(pricing_error-model_estimate))
# dplyr::mutate_(prediction_error=lazyeval::interp(~(abs(pricing_error-model_estimate)/pmax(abs(pricing_error),0.01))))
#Plot
if(export_plots){
if(!dir.exists("learnfin_plots")){
dir.create("learnfin_plots", showWarnings = TRUE, recursive = TRUE, mode = "0777")
}
# data_mesh <-
# expand.grid(scaled_moneyness=0:100,scaled_maturity=0:100) %>%
# dplyr::rowwise() %>%
# dplyr::mutate(clusters=get_the_cluster(scaled_moneyness,scaled_maturity,kmeans_centers)) %>%
# dplyr::ungroup() %>%
# dplyr::left_join(.,kmeans_center_errors,by="clusters")
plot_export_kmeans(result_data)
}
# summary<-
# result_data %>%
# dplyr::group_by(t_or_p) %>%
# dplyr::summarise(prediction_error=median(prediction_error,na.rm=TRUE)) %>%
# dplyr::select(prediction_error) %>%
# unlist
# names(summary)<-c("Prediction","Training")
# return(list(result_data=result_data,summary=summary))
return(result_data)
}
#' @title Plot K-Means
#'
#' @description
#' Plot the Kmeans graphs.
plot_export_kmeans<-function(the_df){
opt_type <- the_df$type[1]
timestamp <- format(Sys.time(),"%Y%m%d%H%M")
#General theme options
plot_theme_opts <- ggplot2::scale_color_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish)
ellipse_display <- ggplot2::stat_ellipse(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),geom="polygon",level=0.999,ggplot2::aes(group=model_estimate,fill=model_estimate),alpha=0.2,linetype=0,show.legend=FALSE)
fill_opts <-
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none")
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ellipse_display +
fill_opts +
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title="Training Data Pricing Errors (K-Means Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_training_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ellipse_display +
fill_opts +
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title="Prediction Data Pricing Errors (K-Means Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_prediction_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot() +
ggplot2::stat_ellipse(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),geom="polygon",level=0.999,ggplot2::aes(x=moneyness,y=maturity,group=model_estimate,fill=model_estimate),alpha=0.4,linetype=0,show.legend=FALSE) +
fill_opts +
ggplot2::guides(fill="colourbar") +
ggplot2::labs(fill="",title=paste0("Cluster Estimates (K-Means Method)")) +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_cluster_estimate_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title="Training Data Prediction Error (K-Means Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_training_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title="Prediction Data Prediction Error (K-Means Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_prediction_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
}
#' @title Manual Learn
#'
#' @description
#' Given a set of moneyness and maturity boundaries and a option data set, this function calculates in-sample and out-of-sample clustering errors.
#' @param raw_data The option data set given in the format of \link{uslfin_ds_1}.
#' @param CallPut It denotes whether to use the call or put options.
#' @param moneyness_breaks Vector of moneyness boundaries.
#' @param moneyness_breaks Vector of maturity boundaries.
#' @export
manual_learn<-function(raw_data,CallPut,moneyness_breaks=c(0.899,0.94,0.97,1.00,1.03,1.06,1.101),maturity_breaks=c(0,42,126,252),export_plots=FALSE){
#First define the manual clusters
transformed_data <-
raw_data %>%
dplyr::filter(type==CallPut) %>%
dplyr::mutate( moneyness_limits=cut(moneyness,moneyness_breaks),
maturity_limits=cut(maturity,maturity_breaks))
#Get the average error in each cluster
manual_centers <-
transformed_data %>%
dplyr::filter(t_or_p=="Training") %>%
dplyr::group_by(moneyness_limits,maturity_limits) %>%
dplyr::summarise(model_estimate=mean(pricing_error,na.rm=TRUE))
#Calculate the prediction error
result_data <-
transformed_data %>%
dplyr::left_join(.,manual_centers,by=c("moneyness_limits","maturity_limits")) %>%
dplyr::mutate(prediction_error=abs(pricing_error-model_estimate))
# dplyr::mutate_(prediction_error=lazyeval::interp(~(abs(pricing_error-model_estimate)/pmax(abs(pricing_error),0.01))))
#Plot
if(export_plots){
if(!dir.exists("learnfin_plots")){
dir.create("learnfin_plots", showWarnings = TRUE, recursive = TRUE, mode = "0777")
}
plot_export_manual(result_data,moneyness_breaks=moneyness_breaks,maturity_breaks=maturity_breaks)
}
# summary<- result_data %>% group_by(TrainPred) %>% summarise(MAPE=mean(MAPE,na.rm=TRUE)) %>% select(MAPE) %>% unlist
# names(summary)<-c("Prediction","Training")
# return(list(result_data=result_data,summary=summary))
return(result_data)
}
#' @title Plot Export (Manual Method)
#'
#' @description
#' Explanatory plots for manual method.
plot_export_manual<-function(the_df,moneyness_breaks,maturity_breaks){
opt_type <- the_df$type[1]
timestamp <- format(Sys.time(),"%Y%m%d%H%M")
#General theme options
plot_theme_opts <- ggplot2::scale_color_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish)
#Calculate moneyness-maturity group corners
mon_range<-data.frame(mon_min=moneyness_breaks[-length(moneyness_breaks)],mon_max=moneyness_breaks[-1])
mat_range<-data.frame(mat_min=maturity_breaks[-length(maturity_breaks)],mat_max=maturity_breaks[-1])
mon_mat_range<-data.frame(mon_min=numeric(),mon_max=numeric(),mat_min=numeric(),mat_max=numeric())
for(i in 1:nrow(mat_range)){
suppressWarnings(mon_mat_range<-rbind(mon_mat_range,cbind(mon_range,mat_range[i,])))
}
mon_mat_range <-
mon_mat_range %>%
dplyr::mutate(mon_mid=(mon_min+mon_max)/2,mat_mid=(mat_min+mat_max)/2) %>%
dplyr::mutate( moneyness_limits=cut(mon_mid,moneyness_breaks),
maturity_limits=cut(mat_mid,maturity_breaks)) %>%
dplyr::left_join(.,the_df %>%
dplyr::distinct(moneyness_limits,maturity_limits,model_estimate), by=c("moneyness_limits","maturity_limits"))
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
#Shaded area plot
ggplot2::geom_rect(data=mon_mat_range,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.25) +
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none") +
#Geom point plot
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Training Data Pricing Errors (Manual Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_training_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
#Shaded area plot
ggplot2::geom_rect(data=mon_mat_range,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.25) +
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none") +
#Geom point plot
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Prediction Data Pricing Errors (Manual Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_prediction_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot() +
ggplot2::geom_rect(data=mon_mat_range,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.75) +
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none") +
ggplot2::labs(fill="",title=paste0("Cluster Estimates (Manual Method)")) +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_cluster_estimate_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Training Data Prediction Error (Manual Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_training_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Prediction Data Prediction Error (Manual Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_prediction_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
}
#' @title DM Learn
#'
#' @description
#' Implements different data mining algorithms (SVM, CIT and DT).
#' @param raw_data The option data set given in the format of \link{uslfin_ds_1}.
#' @param model_name The name of the data mining model to use. Currently there are three options: Support Vector Machine (\code{"svm"}) of \code{e1071} package, Decision Tree (\code{"dt"}) of \code{rpart} package and Conditional Inference Tree \code{("cit")} of the \code{partykit} package.
#' @param CallPut It denotes whether to use the call or put options.
#' @param randseed To set the randomness seed to a known value. Good for reproducibility.
#' @param ... Parameters to pass to data mining algorithm functions. (See individual functions for further explanations.)
#' @export
dm_learn<-function(raw_data,model_name="svm",CallPut="call",randseed=0,export_plots=FALSE,...){
if(randseed>0){
set.seed(randseed)
}
transformed_data <- dplyr::filter(raw_data,type==CallPut)
# training_data<-filter(transformed_data,TrainPred=="Training")
the_formula<- as.formula(paste0("pricing_error"," ~ ","moneyness + maturity"))
#Support vector machine
if(model_name=="svm"){
the_model<-e1071::svm(formula=the_formula,data=transformed_data,subset=(transformed_data$t_or_p=="Training"),...)
}else if(model_name=="cit"){
#Conditional inference tree
the_model<-partykit::ctree(the_formula, data=transformed_data,subset=(transformed_data$t_or_p=="Training"),...)
# print(unlist(the_model$info$control[1:7]))
}else if(model_name=="dt"){
#Decision tree
the_model<-rpart::rpart(the_formula, data=transformed_data,subset=(transformed_data$t_or_p=="Training"),...)
the_model <- partykit::as.party(the_model)
}else{
stop("Wrong model name!")
}
result_data <-
transformed_data %>%
dplyr::mutate(model_estimate=predict(the_model,.)) %>%
dplyr::mutate(prediction_error=abs(pricing_error-model_estimate))
# dplyr::mutate_(prediction_error=lazyeval::interp(~(abs(pricing_error-model_estimate)/pmax(abs(pricing_error),0.01))))
#Plot
if(export_plots & model_name %in% c("dt","cit")){
if(!dir.exists("learnfin_plots")){
dir.create("learnfin_plots", showWarnings = TRUE, recursive = TRUE, mode = "0777")
}
data_mesh <-
expand.grid(seq(from=min(result_data$maturity),to=max(result_data$maturity),length.out=1001),
seq(from=min(result_data$moneyness),to=max(result_data$moneyness),length.out=1001)) %>%
dplyr::rename(maturity=Var1,moneyness=Var2) %>%
dplyr::mutate(node=predict(the_model,.,type = "node")) %>%
dplyr::mutate(model_estimate=predict(the_model,.)) %>%
dplyr::group_by(node) %>%
dplyr::summarise(mon_min=min(moneyness),mon_max=max(moneyness),mat_min=min(maturity),mat_max=max(maturity),model_estimate=mean(model_estimate)) %>%
dplyr::ungroup() %>%
dplyr::mutate(mon_mid=(mon_min+mon_max)/2,mat_mid=(mat_min+mat_max)/2) %>%
dplyr::tbl_df()
plot_export_dm(result_data,data_mesh,model_name)
}
# summary<- result_data %>% group_by(TrainPred) %>% summarise(MAPE=mean(MAPE,na.rm=TRUE)) %>% select(MAPE) %>% unlist
# names(summary)<-c("Prediction","Training")
# return(list(result_data=result_data,summary=summary))
return(result_data)
}
plot_export_dm<-function(the_df,data_mesh,model_name){
opt_type <- the_df$type[1]
timestamp <- format(Sys.time(),"%Y%m%d%H%M")
plot_theme_opts <- ggplot2::scale_color_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish)
plot_rect<-
ggplot2::geom_rect(data=data_mesh,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.25)
fill_opts <-
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none")
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
plot_rect +
fill_opts+
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title=paste0("Training Data Pricing Errors (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_training_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
plot_rect +
fill_opts+
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title=paste0("Prediction Data Pricing Errors (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_prediction_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot() + ggplot2::geom_rect(data=data_mesh,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.75) +
fill_opts +
ggplot2::guides(fill="colourbar") +
ggplot2::labs(fill="",title=paste0("Cluster Estimates (",toupper(model_name)," Method)")) +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_cluster_estimate_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(desc(pricing_error)) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title=paste0("Training Data Prediction Error (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_training_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(desc(pricing_error)) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title=paste0("Prediction Data Prediction Error (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_prediction_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
}
learnfin/learnfin documentation built on May 20, 2019, 11:29 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.
Defines functions kmeans_learn plot_export_kmeans manual_learn plot_export_manual dm_learn
Documented in dm_learn kmeans_learn manual_learn plot_export_kmeans plot_export_manual
#Below functions are individual DM algorithms and other supplementary functions used in uslfin package
#' @title K-Means Learn
#'
#' @description
#' Implements a specialized version of K-Means algorithm on the data set. When creating the clusters (in-sample) the function uses pricing_error, moneyness and maturity. But when predicting, it uses only moneyness and maturity covariates. All covariates are scaled between 0-100.
#'
#' @param raw_data The option data set given in the format of \link{uslfin_ds_1}.
#' @param CallPut It denotes whether to use the call or put options.
#' @param randseed To set the randomness seed to a known value. Good for reproducibility.
#' @param moneyness_interval Minimum and maximum of moneyness values. Required for rescaling.
#' @param maturity_interval Minimum and maximum of maturity values. Required for rescaling. In trading days (one year \= 252 days)
#' @param n_cluster Number of clusters. If \code{0}, then number of clusers is determined by the contract set size. Exact calculation is \code{min(round(nrow(training_matrix)/100),200)}
#' @export
kmeans_learn<-function(raw_data,CallPut="call",randseed=0,moneyness_interval=c(0.9,1.1),maturity_interval=c(4,252),n_cluster=0,export_plots=FALSE){
scaled_data <-
raw_data %>%
#Use only appropriate option type (call or put)
dplyr::filter(type == CallPut) %>%
#Rescale moneyness and maturity to a 0-100 scale each.
dplyr::mutate_(scaled_moneyness = lazyeval::interp(~((moneyness-min(moneyness_interval))/(max(moneyness_interval)-min(moneyness_interval))*100)),
scaled_maturity = lazyeval::interp(~((maturity-min(maturity_interval))/(max(maturity_interval)-min(maturity_interval))*100)))
#Prepare the training data
training_data <-
scaled_data %>%
#Filter only training data
dplyr::filter(t_or_p == "Training") %>%
dplyr::mutate_(scaled_error=lazyeval::interp(~(rescale_error(pricing_error,mederror=NA)*100)))
#Create the error + moneyness + maturity matrix for kmeans
training_matrix<-
training_data %>%
dplyr::select(scaled_error,scaled_moneyness,scaled_maturity) %>%
as.matrix
#Number of clusters are determined by the number of contracts (n_contracts/100). Max 200 clusters
if(n_cluster == 0){
n_cluster <- min(round(nrow(training_matrix)/100),200)
}
#Set the seed if there is a predetermined one
if(randseed>0){
set.seed(randseed)
}
#Train K-Means
kmeans_training <- kmeans(training_matrix,centers=n_cluster,nstart=round(n_cluster/5),iter.max=10^6)
#Get the cluster centers
kmeans_centers <- kmeans_training$centers[,c("scaled_moneyness","scaled_maturity")]
#Get the average pricing error of each cluster
kmeans_center_errors <-
training_data %>%
dplyr::mutate(clusters=kmeans_training$cluster) %>%
dplyr::group_by(clusters) %>%
dplyr::summarise_(model_estimate=lazyeval::interp(~mean(pricing_error,na.rm=TRUE)))
#Get both training and prediction contracts' prediction error on pricing errors.
result_data <-
scaled_data %>%
dplyr::rowwise() %>%
dplyr::mutate(clusters=get_the_cluster(scaled_moneyness,scaled_maturity,kmeans_centers)) %>%
dplyr::ungroup() %>%
dplyr::left_join(.,kmeans_center_errors,by="clusters") %>%
dplyr::mutate(prediction_error=abs(pricing_error-model_estimate))
# dplyr::mutate_(prediction_error=lazyeval::interp(~(abs(pricing_error-model_estimate)/pmax(abs(pricing_error),0.01))))
#Plot
if(export_plots){
if(!dir.exists("learnfin_plots")){
dir.create("learnfin_plots", showWarnings = TRUE, recursive = TRUE, mode = "0777")
}
# data_mesh <-
# expand.grid(scaled_moneyness=0:100,scaled_maturity=0:100) %>%
# dplyr::rowwise() %>%
# dplyr::mutate(clusters=get_the_cluster(scaled_moneyness,scaled_maturity,kmeans_centers)) %>%
# dplyr::ungroup() %>%
# dplyr::left_join(.,kmeans_center_errors,by="clusters")
plot_export_kmeans(result_data)
}
# summary<-
# result_data %>%
# dplyr::group_by(t_or_p) %>%
# dplyr::summarise(prediction_error=median(prediction_error,na.rm=TRUE)) %>%
# dplyr::select(prediction_error) %>%
# unlist
# names(summary)<-c("Prediction","Training")
# return(list(result_data=result_data,summary=summary))
return(result_data)
}
#' @title Plot K-Means
#'
#' @description
#' Plot the Kmeans graphs.
plot_export_kmeans<-function(the_df){
opt_type <- the_df$type[1]
timestamp <- format(Sys.time(),"%Y%m%d%H%M")
#General theme options
plot_theme_opts <- ggplot2::scale_color_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish)
ellipse_display <- ggplot2::stat_ellipse(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),geom="polygon",level=0.999,ggplot2::aes(group=model_estimate,fill=model_estimate),alpha=0.2,linetype=0,show.legend=FALSE)
fill_opts <-
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none")
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ellipse_display +
fill_opts +
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title="Training Data Pricing Errors (K-Means Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_training_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ellipse_display +
fill_opts +
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title="Prediction Data Pricing Errors (K-Means Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_prediction_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot() +
ggplot2::stat_ellipse(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),geom="polygon",level=0.999,ggplot2::aes(x=moneyness,y=maturity,group=model_estimate,fill=model_estimate),alpha=0.4,linetype=0,show.legend=FALSE) +
fill_opts +
ggplot2::guides(fill="colourbar") +
ggplot2::labs(fill="",title=paste0("Cluster Estimates (K-Means Method)")) +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_cluster_estimate_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title="Training Data Prediction Error (K-Means Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_training_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title="Prediction Data Prediction Error (K-Means Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/kmeans_method_prediction_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
}
#' @title Manual Learn
#'
#' @description
#' Given a set of moneyness and maturity boundaries and a option data set, this function calculates in-sample and out-of-sample clustering errors.
#' @param raw_data The option data set given in the format of \link{uslfin_ds_1}.
#' @param CallPut It denotes whether to use the call or put options.
#' @param moneyness_breaks Vector of moneyness boundaries.
#' @param moneyness_breaks Vector of maturity boundaries.
#' @export
manual_learn<-function(raw_data,CallPut,moneyness_breaks=c(0.899,0.94,0.97,1.00,1.03,1.06,1.101),maturity_breaks=c(0,42,126,252),export_plots=FALSE){
#First define the manual clusters
transformed_data <-
raw_data %>%
dplyr::filter(type==CallPut) %>%
dplyr::mutate( moneyness_limits=cut(moneyness,moneyness_breaks),
maturity_limits=cut(maturity,maturity_breaks))
#Get the average error in each cluster
manual_centers <-
transformed_data %>%
dplyr::filter(t_or_p=="Training") %>%
dplyr::group_by(moneyness_limits,maturity_limits) %>%
dplyr::summarise(model_estimate=mean(pricing_error,na.rm=TRUE))
#Calculate the prediction error
result_data <-
transformed_data %>%
dplyr::left_join(.,manual_centers,by=c("moneyness_limits","maturity_limits")) %>%
dplyr::mutate(prediction_error=abs(pricing_error-model_estimate))
# dplyr::mutate_(prediction_error=lazyeval::interp(~(abs(pricing_error-model_estimate)/pmax(abs(pricing_error),0.01))))
#Plot
if(export_plots){
if(!dir.exists("learnfin_plots")){
dir.create("learnfin_plots", showWarnings = TRUE, recursive = TRUE, mode = "0777")
}
plot_export_manual(result_data,moneyness_breaks=moneyness_breaks,maturity_breaks=maturity_breaks)
}
# summary<- result_data %>% group_by(TrainPred) %>% summarise(MAPE=mean(MAPE,na.rm=TRUE)) %>% select(MAPE) %>% unlist
# names(summary)<-c("Prediction","Training")
# return(list(result_data=result_data,summary=summary))
return(result_data)
}
#' @title Plot Export (Manual Method)
#'
#' @description
#' Explanatory plots for manual method.
plot_export_manual<-function(the_df,moneyness_breaks,maturity_breaks){
opt_type <- the_df$type[1]
timestamp <- format(Sys.time(),"%Y%m%d%H%M")
#General theme options
plot_theme_opts <- ggplot2::scale_color_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish)
#Calculate moneyness-maturity group corners
mon_range<-data.frame(mon_min=moneyness_breaks[-length(moneyness_breaks)],mon_max=moneyness_breaks[-1])
mat_range<-data.frame(mat_min=maturity_breaks[-length(maturity_breaks)],mat_max=maturity_breaks[-1])
mon_mat_range<-data.frame(mon_min=numeric(),mon_max=numeric(),mat_min=numeric(),mat_max=numeric())
for(i in 1:nrow(mat_range)){
suppressWarnings(mon_mat_range<-rbind(mon_mat_range,cbind(mon_range,mat_range[i,])))
}
mon_mat_range <-
mon_mat_range %>%
dplyr::mutate(mon_mid=(mon_min+mon_max)/2,mat_mid=(mat_min+mat_max)/2) %>%
dplyr::mutate( moneyness_limits=cut(mon_mid,moneyness_breaks),
maturity_limits=cut(mat_mid,maturity_breaks)) %>%
dplyr::left_join(.,the_df %>%
dplyr::distinct(moneyness_limits,maturity_limits,model_estimate), by=c("moneyness_limits","maturity_limits"))
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
#Shaded area plot
ggplot2::geom_rect(data=mon_mat_range,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.25) +
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none") +
#Geom point plot
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Training Data Pricing Errors (Manual Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_training_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
#Shaded area plot
ggplot2::geom_rect(data=mon_mat_range,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.25) +
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none") +
#Geom point plot
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Prediction Data Pricing Errors (Manual Method)",color="") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_prediction_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot() +
ggplot2::geom_rect(data=mon_mat_range,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.75) +
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none") +
ggplot2::labs(fill="",title=paste0("Cluster Estimates (Manual Method)")) +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_cluster_estimate_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Training Data Prediction Error (Manual Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_training_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::geom_vline(xintercept=as.numeric(moneyness_breaks)) +
ggplot2::geom_hline(yintercept=as.numeric(maturity_breaks)) +
ggplot2::labs(title="Prediction Data Prediction Error (Manual Method)",color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/manual_method_prediction_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
}
#' @title DM Learn
#'
#' @description
#' Implements different data mining algorithms (SVM, CIT and DT).
#' @param raw_data The option data set given in the format of \link{uslfin_ds_1}.
#' @param model_name The name of the data mining model to use. Currently there are three options: Support Vector Machine (\code{"svm"}) of \code{e1071} package, Decision Tree (\code{"dt"}) of \code{rpart} package and Conditional Inference Tree \code{("cit")} of the \code{partykit} package.
#' @param CallPut It denotes whether to use the call or put options.
#' @param randseed To set the randomness seed to a known value. Good for reproducibility.
#' @param ... Parameters to pass to data mining algorithm functions. (See individual functions for further explanations.)
#' @export
dm_learn<-function(raw_data,model_name="svm",CallPut="call",randseed=0,export_plots=FALSE,...){
if(randseed>0){
set.seed(randseed)
}
transformed_data <- dplyr::filter(raw_data,type==CallPut)
# training_data<-filter(transformed_data,TrainPred=="Training")
the_formula<- as.formula(paste0("pricing_error"," ~ ","moneyness + maturity"))
#Support vector machine
if(model_name=="svm"){
the_model<-e1071::svm(formula=the_formula,data=transformed_data,subset=(transformed_data$t_or_p=="Training"),...)
}else if(model_name=="cit"){
#Conditional inference tree
the_model<-partykit::ctree(the_formula, data=transformed_data,subset=(transformed_data$t_or_p=="Training"),...)
# print(unlist(the_model$info$control[1:7]))
}else if(model_name=="dt"){
#Decision tree
the_model<-rpart::rpart(the_formula, data=transformed_data,subset=(transformed_data$t_or_p=="Training"),...)
the_model <- partykit::as.party(the_model)
}else{
stop("Wrong model name!")
}
result_data <-
transformed_data %>%
dplyr::mutate(model_estimate=predict(the_model,.)) %>%
dplyr::mutate(prediction_error=abs(pricing_error-model_estimate))
# dplyr::mutate_(prediction_error=lazyeval::interp(~(abs(pricing_error-model_estimate)/pmax(abs(pricing_error),0.01))))
#Plot
if(export_plots & model_name %in% c("dt","cit")){
if(!dir.exists("learnfin_plots")){
dir.create("learnfin_plots", showWarnings = TRUE, recursive = TRUE, mode = "0777")
}
data_mesh <-
expand.grid(seq(from=min(result_data$maturity),to=max(result_data$maturity),length.out=1001),
seq(from=min(result_data$moneyness),to=max(result_data$moneyness),length.out=1001)) %>%
dplyr::rename(maturity=Var1,moneyness=Var2) %>%
dplyr::mutate(node=predict(the_model,.,type = "node")) %>%
dplyr::mutate(model_estimate=predict(the_model,.)) %>%
dplyr::group_by(node) %>%
dplyr::summarise(mon_min=min(moneyness),mon_max=max(moneyness),mat_min=min(maturity),mat_max=max(maturity),model_estimate=mean(model_estimate)) %>%
dplyr::ungroup() %>%
dplyr::mutate(mon_mid=(mon_min+mon_max)/2,mat_mid=(mat_min+mat_max)/2) %>%
dplyr::tbl_df()
plot_export_dm(result_data,data_mesh,model_name)
}
# summary<- result_data %>% group_by(TrainPred) %>% summarise(MAPE=mean(MAPE,na.rm=TRUE)) %>% select(MAPE) %>% unlist
# names(summary)<-c("Prediction","Training")
# return(list(result_data=result_data,summary=summary))
return(result_data)
}
plot_export_dm<-function(the_df,data_mesh,model_name){
opt_type <- the_df$type[1]
timestamp <- format(Sys.time(),"%Y%m%d%H%M")
plot_theme_opts <- ggplot2::scale_color_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish)
plot_rect<-
ggplot2::geom_rect(data=data_mesh,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.25)
fill_opts <-
ggplot2::scale_fill_gradientn(colors=c("green","yellow","red","purple","black"),values=scales::rescale(c(0,10,25,50,100),to=c(0,1),from=c(0,100)),limits=c(0,100),oob=scales::squish,guide="none")
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
plot_rect +
fill_opts+
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title=paste0("Training Data Pricing Errors (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_training_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot <-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(pricing_error) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
plot_rect +
fill_opts+
ggplot2::geom_point(ggplot2::aes(color=pricing_error),alpha=0.75,size=2) +
ggplot2::labs(title=paste0("Prediction Data Pricing Errors (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_prediction_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot() + ggplot2::geom_rect(data=data_mesh,ggplot2::aes(x=mon_mid,y=mat_mid,xmin=mon_min,xmax=mon_max,ymin=mat_min,ymax=mat_max,fill=model_estimate),alpha=0.75) +
fill_opts +
ggplot2::guides(fill="colourbar") +
ggplot2::labs(fill="",title=paste0("Cluster Estimates (",toupper(model_name)," Method)")) +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_cluster_estimate_pricing_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(desc(pricing_error)) %>% dplyr::filter(t_or_p=="Training"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title=paste0("Training Data Prediction Error (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_training_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
the_plot<-
ggplot2::ggplot(data=the_df %>% dplyr::arrange(desc(pricing_error)) %>% dplyr::filter(t_or_p=="Prediction"),mapping=ggplot2::aes(x=moneyness,y=maturity)) +
ggplot2::geom_point(ggplot2::aes(color=prediction_error),alpha=0.7,size=2) +
ggplot2::labs(title=paste0("Prediction Data Prediction Error (",toupper(model_name)," Method)"),color="APE") +
plot_theme_opts +
ggplot2::theme_bw()
ggplot2::ggsave(paste0("learnfin_plots/",model_name,"_method_prediction_prediction_error_",opt_type,"_",timestamp,".png"),the_plot,width=12,height=5)
}
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.