R语言定价-A2variableAnalysis

## single variable analysis



woe_decile_qrt <- data.table()
result_qrt <- data.table()
for (name in names(pricing.merge_data)) {
  print(name)
  for (date in  unique(pricing.merge_data$decision_qrt)) {
    
  
  for(t in unique(pricing.merge_data$group_binning)){
    tmp_tb <- pricing.merge_data[ group_binning == t & decision_qrt == date]
    # tmp_tb <- model.data
    N <- nrow(tmp_tb)
    miss_cnt <- nrow(tmp_tb[(is.na(get(name)) | get(name) == 'null')])
    miss_pct <- miss_cnt/N
    if(length(unique(tmp_tb[,get(name)])) > 1  & !name %in% c('transport_id', 'code_risk_policy') ){
      print(name)
      # try(result <- data.table(name, t, miss_cnt, WoE_Cal(tmp_tb[target >= 0], event = 'target', segVar = name, cut_num = 5, seg = 5)), silent = T )
      t <- try(result_qrt <- data.table(name, t, date,miss_cnt, WoE_Cal(tmp_tb[mob6_cur_dpd30_flag >= 0 & !is.na(get(name))], event = 'mob6_cur_dpd30_flag', segVar = name, cut_num = 5, seg = 5)), silent = T )
      # t <- try(result <- data.table(name, miss_cnt, WoE_Cal(tmp_tb[target >= 0 & !is.na(get(name))], event = 'target', segVar = name, cut_num = 5, seg = 5)), silent = T )
      if(inherits(t, 'try-error')) next
      woe_decile_qrt <- rbind(woe_decile_qrt, result_qrt)
      # if(is.numeric(tmp2[!is.na(get(name)), get(name)])){
      #   try(tmp2[, paste0('release_seg.', name) := cut(age, breaks = paste0(0, unique(tmp2[, get(name)]), Inf))])
      # }else{
      #   try(tmp2[, paste0('release_seg.', name) := get(name)])
      # }
    }
  }
  }
}

woe_decile_qrt[, ks := abs(ks)]
iv_sum_qrt <- woe_decile_qrt[, .(iv = round(sum(ifelse(IV == Inf, 0.0, IV)), digits = 4)
                         , ks = max(ks)
                         , miss_pct =max(miss_cnt)/sum(Total)
                         , cnt = sum(Total)
)
, by = c('name', 't', 'date')]

out <- dcast(iv_sum, t + name ~ date, value.var = 'iv')
write.csv(woe_decile_qrt, 'woe_decile_qrt.csv')


woe_decile_all <- data.table()
result_all <- data.table()
for (name in names(pricing.merge_data)) {
  print(name)
    
    for(t in unique(pricing.merge_data$group_binning)){
      tmp_tb <- pricing.merge_data[ group_binning == t ]
      # tmp_tb <- model.data
      N <- nrow(tmp_tb)
      miss_cnt <- nrow(tmp_tb[(is.na(get(name)) | get(name) == 'null')])
      miss_pct <- miss_cnt/N
      if(length(unique(tmp_tb[,get(name)])) > 1  & !name %in% c('transport_id', 'code_risk_policy') ){
        print(name)
        # try(result <- data.table(name, t, miss_cnt, WoE_Cal(tmp_tb[target >= 0], event = 'target', segVar = name, cut_num = 5, seg = 5)), silent = T )
        t <- try(result_all <- data.table(name, t, miss_cnt, WoE_Cal(tmp_tb[mob6_ever_dpd30_flag >= 0 & !is.na(get(name))], event = 'mob6_ever_dpd30_flag', segVar = name, cut_num = 5, seg = 5)), silent = T )
        # t <- try(result <- data.table(name, miss_cnt, WoE_Cal(tmp_tb[target >= 0 & !is.na(get(name))], event = 'target', segVar = name, cut_num = 5, seg = 5)), silent = T )
        if(inherits(t, 'try-error')) next
        woe_decile_all <- rbind(woe_decile_all, result_all)
        # if(is.numeric(tmp2[!is.na(get(name)), get(name)])){
        #   try(tmp2[, paste0('release_seg.', name) := cut(age, breaks = paste0(0, unique(tmp2[, get(name)]), Inf))])
        # }else{
        #   try(tmp2[, paste0('release_seg.', name) := get(name)])
        # }
      }
    }
  }

woe_decile_all[, ks := abs(ks)]
iv_sum_all <- woe_decile_all[, .(iv = round(sum(ifelse(IV == Inf, 0.0, IV)), digits = 4)
                             , ks = max(ks)
                             , miss_pct =max(miss_cnt)/sum(Total)
                             , cnt = sum(Total)
)
, by = c('name', 't')]


t1 <- dcast(iv_sum_qrt, t + name ~ date, value.var ='iv')

t2 <- iv_sum_all[, c('t', 'name', 'iv'), with = F]

tt <- merge(t2, t1, by = c('t', 'name') )
write.csv(tt, 'out.csv')