Smartloan: A Risk-Aware And Explainable Loan Eligibility Prediction System Using Machine Learning

Abstract

Loan approval is a critical process in banking that demands precision, fairness, and transparency, yet most existing systems still rely on rigid rule-based or basic machine learning models that only classify applications as approved or rejected without explaining the reasoning behind their decisions. To address these limitations, SmartLoan was developed as an intelligent and transparent loan screening system designed to make the approval process more interpretable and stable. Using regional, professional, and financial data, SmartLoan not only predicts loan eligibility but also generates an eligibility score and recommends a safe loan amount tailored to each applicant. The system integrates powerful algorithms such as Random Forest, Logistic Regression, XGBoost, Cosine Similarity, and SHAP (Shapley Additive Explanations) to achieve both accuracy and explainability, comparing each applicant’s profile with similar financial cases to ensure fairness and consistency. By producing clear, data-driven, and human-understandable results, SmartLoan empowers both banks and applicants to comprehend why a decision was made, thereby improving transparency, enhancing decision confidence, and strengthening trust between financial institutions and borrowers. Future extensions may include automated document verification, income trend monitoring, and real-time financial analysis to further enhance reliability and adaptability in dynamic loan approval environments. The SmartLoan system demonstrated excellent performance across multiple models, with XGBoost achieving the highest accuracy of 99.36%, followed by Logistic Regression at 98.78%, Random Forest at 98.50%, and the SmartLoan Ensemble Model (with Cosine Similarity and SHAP integration) achieving 99.07%. These results confirm that SmartLoan not only delivers highly accurate predictions but also maintains fairness and interpretability through its explainable AI framework.