R/woereg.R

In billyuanyao/WOECredit:

#' woereg
#'
#' @author
#' Yuan Yao
#' @title
#' regressions and classifications for data frame in credit firld
#' @description
#' apply logistical regression and trees to credit related data frame, to calcualte the credit score
#' @name woereg
#' @param Data data frame with at least two columns
#' @param keeplist : Name of the Independent Variables keept for capping,
#' if missing then for all Independent Variables
#' @param droplist : Name of the Independent Variables keept for capping,
#' if missing then for all Independent Variables
#' @param y : Name of the dependent Variables
#' @param regtype : there are mutilpe regresiion types:
#' logistical regression, with regtype='logit', currently available, which is default
#' linear regression, with regtype='logit', currently under building
#' multinomial logistic regression, with regtype='multi', currently under building
#' ordinal logistic regression, with regtype='ordinal', currently under building
#' @param pvalue : Name of the dependent Variables
#' the p-value cutoff, would be value between 0-1, for example 0.01, 0.05, the default value is 0.01
#' @param n : number of iterations
#' the number of logistical iterations would run, default 3 (later will based on opt p_value)
#'
#' Note: I used glm.R to train mulptple iterations, since every time we use GLM in logistic regression to build credit models,
#' there would be multiple training iterations and it's very hard to do variable keeping/droping each time.
#' Later, I will have more regression types added into this interface, by using param regtype
#'
#' test:(currectly only works for numarical variables, since the charactor variable names changed in GLM coeficient)
#' data is from the Titanic project https://www.kaggle.com/c/titanic/data
#' traindata <- read.csv('train.csv',header=T,na.strings=c(""))
#' Data <- subset(traindata,select=c(2,3,5,6,7,8,10,12))
#' keeplist= c('Fare','Pclass')
#' droplist= c('Sex','Embarked')
#' woereg(Data=Data,keeplist=c('Fare','Pclass'),droplist=c('Sex','Embarked'),y='Survived', regtype='logit', p_value= 0.01, n=3)
#'
#' results:
#'              Estimate Std. Error z value Pr(>|z|)
#'              (Intercept)  3.188881   0.488922   6.522 6.93e-11 ***
#'              Pclass      -1.130074   0.142036  -7.956 1.77e-15 ***
#'              Age         -0.040656   0.006782  -5.994 2.04e-09 ***
#'              Fare         0.003213   0.002326   1.381    0.167


woereg <-function(Data,keeplist=NULL,droplist=NULL,y, regtype='logit', p_value=NULL, n=NULL)
{
  #' drop all variables in the droplist
  if(!is.null(droplist)) {
    Data <- Data[,!(names(Data) %in% droplist)]
  }
  #' this would be updated later, currectly only for numeric vairabls
  nums <- sapply(Data, is.numeric)
  Data<-Data[ , nums]
  #' assign default values
  if(is.null(p_value)) {
    p_value <- 0.01
  }
  if(is.null(n)) {
    n <- 3
  }
  for (i in 1:n){
  #' run GLM model and do a stepwise vairable selection
  f<-paste(as.name(y), "~.")
  #' paste the fomular in to GLM model
  model <- glm(f,family=binomial(link='logit'),data=Data)
  #' get p-value cutoffs
  pvalue <- round(coef(summary(model))[,4],6)
  varsel <- pvalue[which(pvalue <= p_value)]
  #' drop all qualified variables
  drops<-c(droplist,"(Intercept)")
  vallist<- varsel[ !(names(varsel) %in% drops)]
  #' keep all the variables in the keeplist and qualified p-values
  varselname <- unique(c(names(vallist),keeplist))
  #' prepare dataset for the next iteration
  if(!is.null(varselname)) {
    Data<-subset(Data,select=c(y,varselname))
  }
  }
  summary(model)
  return(model)
}