Evaluation of the client for providing a loan

This bachelor's thesis focuses on discriminating between good and bad clients for loan approval using logistic regression and then evaluating the discriminatory ability of the developed model. The practical research uses real consumer credit data from a bank that published their data in a competition on the Kaggle platform. The analytical and modeling phases are conducted in the Python programming language. The research presented in this paper provides insights from a statistical point of view into the problem of evaluating customers for credit. Before modeling, adjustments and transformations were made to the variables to avoid bias in the model results. Emphasis was placed on adjusting the weights due to an imbalance in the dataset, which led to better classification of default clients. The ROC analysis performed on the training data confirmed the acceptable discriminatory ability of the model, which was confirmed on the test dataset. Cross-validation confirmed the result obtained on the test data. Finally, a discussion of the classification table analysis was conducted, presenting important insights into the model's performance. The methods and procedures applied in this thesis can be used to evaluate the probability of default (PD), a key aspect of banking sector regulations such as Basel II standards.