Pemilihan input berbasis AUC

Pemodelan Risiko Kredit di R

Lore Dirick

Manager of Data Science Curriculum at Flatiron School

Kurva ROC untuk 4 model regresi logistik

Tangkapan Layar 2020-06-22 pukul 6.31.53 PM.png

Kurva ROC untuk 4 model regresi logistik

Tangkapan Layar 2020-06-22 pukul 6.31.44 PM.png

Kurva ROC untuk 4 model regresi logistik

Tangkapan Layar 2020-06-22 pukul 6.31.33 PM.png

Pemangkasan berbasis AUC

1) Mulai dengan model berisi semua variabel (di sini, 7) dan hitung AUC

log_model_full <- glm(loan_status ~ loan_amnt + grade + home_ownership + 
                      annual_inc + age + emp_cat + ir_cat, 
                      family = "binomial", data = training_set)

predictions_model_full <- predict(log_model_full, 
                                  newdata = test_set, type ="response")

AUC_model_full <- auc(test_set$loan_status, predictions_model_full)

Area under the curve: 0.6512

2) Bangun 7 model baru, tiap kali hapus satu variabel, lalu buat prediksi PD pada test set

log_1_remove_amnt <- glm(loan_status ~ grade + home_ownership + annual_inc + age + emp_cat + ir_cat,
                         family = "binomial",
                         data = training_set)

log_1_remove_grade <- glm(loan_status ~ loan_amnt + home_ownership + annual_inc + age + emp_cat + ir_cat,
                          family = "binomial",
                          data = training_set)

log_1_remove_home <- glm(loan_status ~ loan_amnt + grade + annual_inc + age + emp_cat + ir_cat,
                         family = "binomial",
                         data = training_set)

pred_1_remove_amnt <- predict(log_1_remove_amnt, newdata = test_set, type = "response")
pred_1_remove_grade <- predict(log_1_remove_grade, newdata = test_set, type = "response")
pred_1_remove_home <- predict(log_1_remove_home, newdata = test_set, type = "response")
...

3) Pertahankan model dengan AUC terbaik (AUC model penuh: 0,6512)

auc(test_set$loan_status, pred_1_remove_amnt)

Area under the curve: 0.6537

auc(test_set$loan_status, pred_1_remove_grade)

Area under the curve: 0.6438

auc(test_set$loan_status, pred_1_remove_home)

Area under the curve: 0.6537

4) Ulangi hingga AUC menurun (secara signifikan)

Ayo berlatih!

Pemodelan Risiko Kredit di R