Model Selection and Validation for Credit Risk Assessment

Through model selection, it was determined that 'model3' is the final model, which includes variables such as residencestate, creditbureauriskassessment, currentbalance, approvedamount, delinquentaccountsinlastyear, accountage, lengthofcredithistory, noofcredithardpulls, and age. Among them, 'delinquentaccountsinlastyear' and 'noofcredithardpulls' have a significant impact on the model. To further validate the model, the caret train function was used for cross-validation, and the results showed that the accuracy of the predictions for the training and test sets was similar (test: 0.8003966; train: 0.8015851). It is worth noting that when comparing the fourfoldplot, the model's accuracy is very high for the dependent variable of 0, but very low for 1.