Indian Journal of Engineering Research Networking and Development (Online ISSN:3048-7676)

The rapid expansion of digital banking has intensified demand for loan approval mechanisms that are simultaneously efficient, accurate, and transparent. Conventional credit evaluation approaches depend on manual assessment and rigid rule-based criteria, which are inherently time-consuming, susceptible to inconsistency, and frequently introduce human bias into financial decisions. This paper presents a machine learning-based framework for predicting bank loan approval outcomes, augmented with Explainable Artificial Intelligence (XAI) to address transparency deficits in automated lending systems. The proposed architecture employs the German Credit structured dataset encompassing applicant attributes such as credit history, income level, employment duration, requested loan amount, and savings status. An ensemble Random Forest classifier is trained to distinguish approved from rejected applications. SHAP and LIME are integrated into the prediction pipeline, furnishing both global feature importance rankings and per-applicant decision rationales. A supplementary rule-based explanation layer translates model logic into human-readable decision steps. Empirical evaluation shows the Random Forest model achieves 78.5% accuracy while maintaining full interpretability.