R语言分箱

#smbinning分箱  两种方式确定切分点 ；1.根据smbinning自带的ctree算法进行分裂，找出最优分割点，然后计算woe和iv
####2.如下计算方式，自己设定切分点，下面是基于数据分布的分位数进行切分

per<-as.vector(quantile(traindata$m03,probs=seq(0,1,0.2),na.rm=T))  
breaks<-per[2:(length(per)-1)]  
result_m03=smbinning.custom(df=traindata,y='y',x="m03",cuts=breaks)  
result_m03$ivtable 



########################## ctree IV
#####
library(smbinning)
br_shenqing_train_d=br_shenqing_train
iv_table=data.frame()
smbin_feature_table=data.frame()

for(i in 1:length(numer_names)){
  br_shenqing_train_d[,numer_names[i]]<-as.numeric(br_shenqing_train_d[,numer_names[i]])
  
  if(sum(is.na(br_shenqing_train_d[,numer_names[i]]))!=10932){
    br_shenqing_train_d[,numer_names[i]]<-as.numeric(br_shenqing_train_d[,numer_names[i]])
    
    ####分箱 
    m=smbinning(df=br_shenqing_train_d,y="target",x=numer_names[i])
    if(m != "No significant splits" && m != "Uniques values < 5"){
      ####保存切分点
      smbin_fm<-m$ivtable[,c('Cutpoint','WoE','IV')]
      names_f=data.frame(numer_names[i],fix.empty.names=FALSE)
      names_f=t(names_f)
      colnames(names_f)<-c('feature_names') 
      smbin_feature_ta<-cbind(names_f,m$ivtable[,c('Cutpoint','WoE','IV')])
      smbin_feature_table<-rbind(smbin_feature_table,smbin_feature_ta)
      
      #####保留feature IV
      iv_ta=data.frame(names=numer_names[i],IV=m$iv)
      iv_table=rbind(iv_table,iv_ta)
    }
    
  }
  
}


iv_table
smbin_feature_table