International Journal of Scientific Research and Engineering Development

Loan eligibility prediction is a critical function in modern banking and financial services that directly impacts institutional risk and customer experience. Manual assessment processes are time-consuming, inconsistent, and susceptible to human bias, making automated machine learning solutions essential. This paper proposes a multi-model ensemble approach for loan eligibility prediction that integrates XGBoost, Random Forest, and Logistic Regression classifiers with a comprehensive feature engineering pipeline. The system processes eight key applicant attributes including credit score, annual income, debt-to-income ratio, employment duration, loan amount, credit history length, number of dependents, and property area. Experiments conducted on a dataset of 45,000 loan applications demonstrate that the proposed XGBoostbased ensemble achieves 97.3% accuracy, 96.9% F1-score, and an AUC-ROC of 0.987, outperforming all baseline models. The system also provides interpretable predictions using SHAP-based feature importance analysis, satisfying regulatory explainability requirements under RBI Fair Practices Code and the EU AI Act.

Harish S G, K.Thenmozhi,"Loan Eligibility Prediction" International Journal of Scientific Research and Engineering Development, V9(2): Page(217-221) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.