Credit risk estimation with machine learning and artifical neural networks algorithms

Abstract

Credit risk assessment is very important for financial institutions today. The probability that a financial institution customer will not be able to repay the credits used is called credit risk. Financial institutions accept or reject credit applications. Institutions evaluate credit applications according to the personal information of the customers, life situation, loyalty, etc. If these data are below various values, financial institutions reject the application. The organization rejected the application because the client anticipated financial difficulties in the future. In the project, "German Credit" data on the Kaggle platform was used. In this data set, customers information and credit status are found as "good" and "bad". By using these data, it is aimed to evaluate new credit application requests. The data set used was passed through various pre-data processing steps and models such as Logistic Regression, Artificial Neural Networks, K-NN, Support Vector, Naïve Bayes, Decision Trees, Random Forest, LGBM and XGB were trained. The highest accuracy is achieved using the XGB model. (0.74)