Prediction of Loan Eligibility Approval

Abstract

The data available through Kaggle includes loan-specific features comprising income level along with credit history as well as asset ownership information and the loan value and education level of the applicant. The main goal consists of creating an accurate forecasting model to categorize loan applications into eligible or ineligible decisions. Seven supervised learning approaches including Decision Tree and Random Forest and Logistic Regression together with SVM and KNN and GNB have been applied for evaluation. SMOTE was used to overcome the unequal distribution of loan status classification labels. The accuracy rate of 98.55% was achieved by Random Forest among the group of models and resulted in superior performance compared to other classifiers regarding precision and recall metrics. The integrated Django web application allows users to provide new data inputs which generate real-time eligibility predictions using the trained model. Such a system represents an efficient and expandable solution designed for institutions trying to automate their loan application review methods.