R语言定价-D1-NCL

collection <- read.csv('E:/03 工作代码/02 R project/风险定价/collection.csv', stringsAsFactors = F)
collection <- data.table(collection)


pricing.result <- merge(collection, t, by = 'transport_id')
pricing.result[, m2_flag := ifelse(overdue_states >= 2, 1, 0)]
pricing.result[, enr_flag := ifelse(overdue_states < 7, 1, 0)]
pricing.result[, gwo_flag := ifelse(overdue_states == 7, 1, 0)]
pricing.result[, mob := 12 * (year(timee) - year()) + month(timee)]
result_sum <- pricing.result[
                             , .(enr_amt = sum(bal_corpus * enr_flag, na.rm = T)
                                  , enr_cnt = sum(enr_flag)
                                  , coin_M2_amt = sum(enr_flag * m2_flag * bal_corpus, na.rm = T)
                                  , coin_M2_cnt = sum(enr_flag * m2_flag, na.rm = T)
                                  , gwo_cnt = sum(gwo_flag)
                                  , gwo_bal = sum(gwo_flag * bal_corpus)
                                  , adj_enr_amt = sum(ifelse(adj_amt_contract == amt_contract, bal_corpus * enr_flag,
                                                        ifelse(bal_corpus * enr_flag > 0, bal_corpus * enr_flag * adj_amt_contract/amt_contract, 0 )), na.rm = T)
                                  , adj_gwo_bal = sum(ifelse(adj_amt_contract == amt_contract, gwo_flag * bal_corpus,
                                                            ifelse(adj_amt_contract < gwo_flag * bal_corpus, adj_amt_contract, gwo_flag * bal_corpus )), na.rm = T)
                                 )
               , by = c('timee', 'platform','group_binning' ,'price_level')]

x1 <- dcast(result_sum, timee + group_binning ~ price_level, value.var = 'coin_M2_cnt')
write.csv(result_sum, 'x1.csv')



########################################################## 最开始的变量统计
stat_table <- data.table()

for(name in names(pricing.merge_data)){
  print(name)
  for(group in unique(pricing.merge_data$group_binning)){
      for(t in unique(pricing.merge_data$decision_qrt)){
        tmp <- pricing.merge_data[flag_loan == 1 & group_binning == group & decision_qrt == t]
        if(is.numeric(tmp[, get(name)])){
          N <- nrow(tmp)
          miss_cnt <- nrow(tmp[(is.na(get(name)) | get(name) == 'null')])
          miss_pct <- miss_cnt/N
          unique_cnt <- length(unique(tmp[, get(name)], na.rm = T))
          maxValue <- max(tmp[, get(name)], na.rm = T)
          minValue <- min(tmp[, get(name)], na.rm = T)
          avgValue <- mean(tmp[, get(name)], na.rm = T)
          P10 <- quantile(tmp[, get(name)], 0.1, na.rm = T)
          P30 <- quantile(pricing.merge_data[, get(name)], 0.3, na.rm = T)
          P50 <- quantile(tmp[, get(name)], 0.5, na.rm = T)
          P70 <- quantile(tmp[, get(name)], 0.7, na.rm = T)
          P90 <- quantile(tmp[, get(name)], 0.9, na.rm = T)
          P99 <- quantile(tmp[, get(name)], 0.99, na.rm = T)
          
          stat_table <- rbind(stat_table, data.table( group, t, name, N, miss_cnt, miss_pct, unique_cnt, maxValue, minValue, avgValue, P10, P30, P50, P70, P90, P99))
          
        }
      }
  }
      
}

write.csv(stat_table, 'variable_profile.csv')