Assessing bank loan applicant (credit score) through ML: a comparative approach

The assessment of credit risk and the appraisal of credit portfolios are of utmost importance to financial institutions that offer loans to businesses and people. A non-performing loan (NPL) refers to a category of loans where the borrower has exhibited delinquency by failing to make the required payments within the designated timeframe. This research examines various machine learning (ML) techniques to address the bank loan defaulter problem. The authors present a comparative analysis of commonly utilized ML models in the Non Performing Loan domain using a dataset from Kaggle. The utilization of class weights addresses the issue of class imbalance having 4269 samples and 13 columns to compare performance metrics by considering Precision, Recall, F1 Score, R-squared values, and Specificity. Based on the performance measures, it can be observed that decision tree and random forest classifiers have 98% accuracy in delivering the most favorable results for the given dataset where decision tree has slightly better result.