Course project on Credit scoring
The goal of this project is to predict loan repayment for Home Credit Default Risk dataset. For that propose we collected old customer data, aggregated them and compared several approaches: linear models with WOE vs ML models like neural networks and boostings. To be more specific: XGBoost, CatBoost, MLP, TabNEt. Finally, the profits for the best models were calculated and the most important features were selected. The resulted product is a web application with a trained XGBoost model (the best score for considered data) on the backend.