A Secure Bank Loan Prediction System by Bridging Differential Privacy and Explainable Machine Learning

• DOI: 10.3390/electronics14081691
• Publication Date: 2025-04-22T00:38:26Z
• Authors: Muhammad Minoar H...; Mohammad Mamun; Arslan Munir; Mohammad Motiur R...; Safiul Haque Chow...
• Keywords: Differential Privacy

Bank loan prediction (BLP) analyzes the financial records of individuals to conclude possible status. Financial always contain confidential information. Hence, privacy is significant in BLP system. This research aims generate a privacy-preserving automated scheme. To achieve this, differential (DP) combined with machine learning (ML). Using benchmark dataset, proposed method two different DP techniques, namely Laplacian and Gaussian, five ML models: Random Forest (RF), Extreme Gradient Boosting (XGBoost), Adaptive (AdaBoost), Logistic Regression (LR), Categorical (CatBoost). Each techniques evaluated by varying distinct parameters 10-fold cross-validation, from outcome analysis, optimal are nominated balance security. The analysis indicates that applying mechanism budget 2 RF model achieves highest accuracy 62.31%. For Gaussian method, best 81.25% attained CatBoost 1.5. Additionally, uses explainable artificial intelligence (XAI) show conclusion capability DP-integrated models. shows an efficient for while preserving personal information and, thus, mitigating vulnerability scams fraud.