A Comprehensive Framework for Predicting Credit Defaults Using Ensemble Methods

In this Work, we why not take care of how to foresee credit hazard in the Home Credit rivalry utilizing machine learning. The objective is to evaluate the likelihood of loan defaults using feature engineering and a model ensemble consisting of LightGBM and CatBoost. The dataset consists of personal, financial and credit bureau attributes, which are converted into useful features. We use stratified k-fold cross-validation to increase the robustness of our models and we apply a voting ensemble approach to obtain a better prediction accuracy. The last model shows a good AUC score indicating that ensemble learning is beneficial for assessing credit risk and to help the lender with informed decision making.