Authors: Ishvik Reddy
Abstract: Traditional credit scoring methods often fail to capture the multi-dimensional complexities of modern financial risks, particularly in volatile markets and for borrowers with limited credit histories. This review article investigates the integration of Machine Learning (ML)-based credit scoring models within the SAP financial and banking ecosystem. We evaluate the transition from legacy logistic regression scorecards to advanced ensemble methods like XGBoost and Random Forests, implemented through the SAP HANA Predictive Analytics Library (PAL) and SAP Business Technology Platform (BTP). The study highlights how the "embedded" and "side-by-side" architectural patterns in SAP S/4HANA enable real-time, data-driven credit decisioning by processing transactional data at the source. Furthermore, the article addresses the critical requirement for Explainable AI (XAI) using SHAP and LIME to meet regulatory standards like Basel IV and GDPR. We explore diverse use cases, including retail loan automation, dynamic corporate credit limit management, and SME financing via alternative data. The study concludes by discussing the future impact of Generative AI and Quantum Machine Learning on credit risk reporting and simulation. By synthesizing technical implementation strategies with financial risk theory, this paper provides a strategic roadmap for banks aiming to deploy transparent, accurate, and high-performance scoring systems within their enterprise landscape.
DOI: http://doi.org/10.5281/zenodo.18229000
