隨機欠抽樣

R 的詐欺偵測

Bart Baesens

Professor Data Science at KU Leuven

隨機欠抽樣（RUS）

類別長條圖：欠抽樣

原始資料：訓練/測試

隨機欠抽樣：步驟 0

隨機欠抽樣：步驟 1

資料集檢視：V2 對 V1

table(creditcard$Class)

    0     1 
24108   492

n_fraud <- 492
new_frac_fraud <- 0.50
new_n_total <- n_fraud / new_frac_fraud ## = 492 / 0.50 = 984

library(ROSE)
undersampling_result <- ovun.sample(formula = Class ~ ., data = creditcard,
                                    method = "under", N = new_n_total, seed = 2018)

undersampled_credit <- undersampling_result$data

prop.table(table(undersampled_credit$Class))

  0   1 
0.5 0.5

欠抽樣後的 V2 對 V1

兩者都試試！

類別長條圖：兩者皆用

n_new <- nrow(creditcard) ## = 24600
fraction_fraud_new <- 0.50

sampling_result <- ovun.sample(formula = Class ~ ., data = creditcard,
                               method = "both",
                               N = n_new,
                               p = fraction_fraud_new,
                               seed = 2018)

sampled_credit <- sampling_result$data

prop.table(table(sampled_credit$Class))

        0         1 
0.5039837 0.4960163

兩者皆用：結果

一起來練習吧！

R 的詐欺偵測