Comparative Study of Credit Score Modelling using Ensemble Techniques

Abstract

Credit scoring, a cornerstone of modern finance, plays a pivotal role in assessing the creditworthiness of individuals and businesses. The primary motivation driving this review paper is to scrutinize the suitability and efficacy of algorithms in the context of credit scoring. This comprehensive review delves into the evolving landscape of credit risk assessment and explores the transformative impact of algorithmic decision making on reducing bias, overfitting and enhancing predictive accuracy. Over time, the accuracy of credit scoring models has seen substantial improvements, reflecting advancements in computational capabilities and data availability. By focusing on a spectrum of algorithms, from classical statistical methods to state-of-the-art machine learning and neural network approaches, this paper provides a comparative analysis of their strengths and limitations in the realm of credit scoring. The discussion encompasses considerations of interpretability, predictive accuracy, scalability, and fairness, shedding light on the trade-offs involved in algorithm selection.