Loan Approval System Using Machine Learning Scoring

Abstract

This Paper Presents LoanAI, A Machine Learning-based Loan Approval System Designed To Address Financial Exclusion Among Young Borrowers Lacking Traditional Credit History. The System Evaluates Creditworthiness Using Bank Transaction Data Such As Salary Consistency, Savings Patterns, And Cash Dependency. An XGBoost Classifier Is Used To Generate A Simulated CIBIL Score Ranging From 300 To 850. A Fraud Detection Mechanism Identifies Anomalies Including High FOIR And Irregular Transaction Patterns. The System Processes Loan Applications In Under 3 Seconds And Improves Approval Rates From 40% To 65% While Maintaining A Low Default Rate Of 4.2%. The Proposed Solution Demonstrates 82% Prediction Accuracy And Enables Scalable, Real-time Loan Decision-making.

Keywords