Intelligent Banking Analytics System for Loan Approval and Customer Churn Prediction Using Machine Learning

Abstract

In the rapidly evolving financial sector, the demand for secure, efficient, and accurate loan approval systems has increased significantly. Traditional processes involve manual verification, subjective decision-making, and time-consuming procedures leading to delays. This paper presents LoanOracle, a machine learning-based prediction and approval system that employs Logistic Regression, Decision Tree, and Random Forest for loan eligibility prediction, and Voting Classifier (GaussianNB + AdaBoost + KNN) and Extra Tree Classifier for customer churn prediction. Models are trained on historical banking datasets evaluating credit history, income, employment status, and repayment capacity. The Django-based web application provides real-time feedback, an admin dashboard for monitoring, and analytical reporting — reducing processing time, eliminating human bias, and enhancing customer satisfaction.

Introduction

The text describes LoanOracle, an intelligent banking analytics system designed to improve loan approval and customer churn prediction using machine learning.

Traditional banking systems rely on manual, slow, and biased processes for loan approvals and lack effective tools to predict customer churn, which affects profitability. To solve this, LoanOracle uses machine learning to automate decision-making and provide real-time predictions.

The system applies multiple ML algorithms such as Logistic Regression, Decision Tree, Random Forest (for loan approval), and Voting Classifier and Extra Tree Classifier (for churn prediction). Among these, Random Forest and ensemble methods perform the best in accuracy and reliability.

LoanOracle is implemented as a full-stack web application using Django (or Flask), Scikit-learn, and a secure database (Supabase/MySQL). It includes modules for user input, data preprocessing, feature selection, loan prediction, churn prediction, risk analysis, secure storage, and admin monitoring.

The system is trained on two datasets: one for loan eligibility and another for customer churn prediction. Data preprocessing includes handling missing values, encoding categorical variables, and balancing class distribution. Models are evaluated using metrics like accuracy, F1-score, precision, and recall.

System architecture follows a three-tier design: frontend (UI), backend (Django ML processing), and database layer. The workflow involves user input → preprocessing → ML model prediction → result display with confidence scores.

Experimental results show strong performance, with Random Forest achieving the highest loan prediction accuracy (~91%) and ensemble models performing well for churn detection. The system provides efficient, scalable, and automated banking decision support.

Conclusion

This paper presented LoanOracle, an intelligent banking analytics system integrating loan approval prediction and customer churn prediction into a unified, deployed web application. By combining Logistic Regression, Decision Tree, Random Forest, Voting Classifier, and Extra Tree Classifier, the system provides accurate, real-time decisions that reduce manual effort, minimise human bias, and enhance efficiency. The modular Django-based architecture with secure database storage and an intuitive admin dashboard positions LoanOracle as a scalable solution for modern banking. Future enhancements: (1) real-time banking API integration for live credit score retrieval; (2) deep learning models for complex financial pattern recognition; (3) fraud detection module using anomaly detection; (4) mobile application for field-level accessibility; (5) AI-powered chatbot for application guidance; (6) cloud-based deployment with big data infrastructure.

References

[1] Ugwuishiwu, C. H. et al. (2022) ‘Machine Learning Models for Predicting Bank Loan Eligibility,’ Intl. Journal of Advanced Trends in Computer Science and Engineering, Vol. 11, No. 2. [2] Majumdar, A. et al. (2022) ‘Loan Prediction using Machine Learning Techniques,’ Intl. Journal of Computer Applications, Vol. 184, No. 25, pp. 1–6. [3] Boddepalli, R. (2022) ‘Loan Approval Prediction System Using Machine Learning,’ Intl. Journal of Engineering Research and Technology, Vol. 11, No. 5, pp. 654–658. [4] Kadam, A. S., Patil, S. and Kulkarni, P. (2021) ‘Prediction for Loan Approval using Machine Learning Algorithm,’ Intl. Journal of Scientific Research in Computer Science, Vol. 9, No. 3, pp. 45–50. [5] Murthy, P. L. S. (2020) ‘Loan Approval Prediction System Using Machine Learning,’ Intl. Journal of Engineering and Technology, Vol. 8, No. 4. [6] Breiman, L. (2001) ‘Random Forests,’ Machine Learning, Vol. 45, No. 1, pp. 5–32. [7] Pedregosa, F. et al. (2011) ‘Scikit-learn: Machine Learning in Python,’ Journal of Machine Learning Research, Vol. 12, pp. 2825–2830. [8] Sundararajan, S. and Ravi, V. (2020) ‘Machine Learning Approach for Credit Risk Assessment,’ Intl. Journal of Advanced Computer Science and Applications, Vol. 11, No. 6, pp. 250–256.

Copyright

Copyright © 2026 Gokul U, Murali S, Anuj B, Ajith A, Mrs. D Ruba. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.