The Use of Machine Learning Adoption in Loan Approval Prediction System

Abstract

Recently, financial institutions have increasingly relied on automated systems to streamline loan approval processes. This study developed a Loan Approval Prediction System using machine learning techniques to evaluate applicant eligibility based on historical loan data. The dataset, sourced from Kaggle, included key attributes such as credit score, income, employment status, loan amount, and debt-to-income ratio. The methodology followed a systematic approach consisting of data preprocessing, exploratory data analysis (EDA), feature selection, target balancing, and model evaluation. Preprocessing involved dropping irrelevant columns, handling missing values, and encoding categorical data, followed by correlation analysis to guide feature selection. To enhance predictive performance, three feature selection techniques—filter, wrapper, and hybrid—were compared using a Support Vector Classifier (SVC). Model evaluation employed accuracy, precision, recall, and F1-score metrics. Results revealed that the hybrid selection method, when combined with Synthetic Minority Over-sampling Technique (SMOTE) and SVC, achieved the highest accuracy (94.93%) and F1-score (95.91%). Observations indicated that applicant income, credit score, and debt-to-income ratio were the most significant predictors of loan approval. The study concludes that the hybrid-SVC model provides an efficient, unbiased, and highly accurate loan approval prediction framework, reducing processing time and decision errors while enhancing transparency and customer satisfaction.