AN IMPROVED APPROACH TO CREDIT RISK ASSESSMENT BASED ON THE RANDOM FOREST ALGORITHM IN CREDIT SCORING SYSTEMS

Abstract

This article analyzes the effectiveness of classical scoring approaches and modern machine learning models used for assessing customer creditworthiness in banking systems. During the study, the performance of Logistic Regression, Decision Tree (J48), Naive Bayes, k-Nearest Neighbor, and Random Forest algorithms in credit scoring tasks was experimentally evaluated using WEKA and KNIME software environments. The experiments were conducted based on the German Credit Data and Default of Credit Card Clients datasets.