Modélisation du risque de crédit en R
Lore Dirick
Manager of Data Science Curriculum at Flatiron School



1) Commencez par un modèle incluant toutes les variables (ici, 7) et calculez l'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$loan_status, predictions_model_full)
Area under the curve: 0.6512
2) Créez 7 nouveaux modèles en retirant à chaque fois une variable, puis faites des prédictions PD sur l'ensemble de test
log_1_remove_amnt <- glm(loan_status ~ grade + home_ownership + annual_inc + age + emp_cat + ir_cat,
family = "binomial",
data = training_set)
log_1_remove_grade <- glm(loan_status ~ loan_amnt + home_ownership + annual_inc + age + emp_cat + ir_cat,
family = "binomial",
data = training_set)
log_1_remove_home <- glm(loan_status ~ loan_amnt + grade + annual_inc + age + emp_cat + ir_cat,
family = "binomial",
data = training_set)
pred_1_remove_amnt <- predict(log_1_remove_amnt, newdata = test_set, type = "response")
pred_1_remove_grade <- predict(log_1_remove_grade, newdata = test_set, type = "response")
pred_1_remove_home <- predict(log_1_remove_home, newdata = test_set, type = "response")
...
3) Conservez le modèle avec la meilleure AUC (AUC du modèle complet : 0,6512)
auc(test_set$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) Répétez jusqu'à ce que l'AUC diminue (significativement)
Modélisation du risque de crédit en R