Vývoj a evaluace skóringových modelů pro bankovní retailový sektor

This thesis is focused on the development and evaluation of scoring models for predicting the risk of client default in the banking retail sector. The processed data were obtained from a publicly available competition organized by the non-bank financial institution Home Credit. Chapters 1 to 3 describe the issue of credit risk and the development of scoring models. Chapter 4 is devoted to the practical development of five models using a classical statistical approach represented by logistic regression, machine learning methods (Random Forest, LightGBM, MLP), and a hybrid model. Based on the evaluation of model performance metrics, the best result was achieved by the hybrid model combining gradient boosting algorithms and logistic regression. The outcomes of the model development are discussed and summarized in the conclusion of the thesis.