Credit Score Classification Using Advanced Machine Learning: A Comprehensive Approach

Abstract

This paper presents a comprehensive machine learning approach for credit score classification, addressing key challenges in financial risk assessment. We propose an optimized CatBoost-based framework that integrates advanced feature engineering, systematic class imbalance handling, and robust evaluation metrics. Our methodology achieves strong classification performance, with AUC scores of 0.944, 0.858, and 0.928 for the Poor, Standard, and Good credit score classes, respectively. The system particularly excels in distinguishing high-risk (Poor) and low-risk (Good) credit profiles, while the Standard class remains the most challenging due to its overlapping characteristics. Through extensive experimentation and analysis, we provide valuable insights into feature importance and model behavior, offering practical implications for financial institutions and credit scoring systems.