Volume 2 - Issue 1, January - February 2026

In the rapidly evolving financial sector, automating credit risk assessment is essential to improve efficiency and reduce default rates. Traditional rule-based systems often fail to handle complex financial data, resulting in delays and inaccurate decisions. This paper presents an Ensemble-Based Loan Approval Prediction System that leverages machine learning to deliver accurate, realtime, and transparent decision support. The study utilizes a dataset of 4,269 records with 12 financial attributes, including income, CIBIL score, and loan terms. A robust preprocessing pipeline incorporating Z-score–based outlier removal, categorical encoding, and a composite Assets feature was implemented. Seven supervised learning algorithms were evaluated, along with three ensemble techniques—Bagging, Random Forest, and AdaBoost—to enhance predictive stability. The Bagging Classifier achieved the best performance, with a testing accuracy and F1-score of 98.36%. To address the interpretability challenge of ensemble models, SHAP (SHapley Additive exPlanations) was integrated to quantify feature contributions, identifying the CIBIL score as the dominant factor in loan approval. The final model is deployed using a Streamlit web application, providing instant predictions and visual explanations for improved decision transparency.

Rohan Kumar Jha, Rohit Kumar, S Hiten,"Ensemble Based Loan Approval Prediction System" International Journal of Scientific Research and Engineering Development, V2(1): Page(337-342) Jan-Feb 2026. ISSN: 3107-6513. www.ijamred.com. Published by Scientific and Academic Research Publishing.