A Comparative Performance Assessment for Prediction of Loan Approval in Financial Sector

Abstract

Loans are the bank’s assets since they generate income in terms of interest to banks. Lending a loan to a customer creates credit and liability for the bank and the customer. The profit and loss of a bank depend on the customer’s ability to pay back the loan or not, i.e., defaulter or not. Therefore, predicting the probability of loan repayment becomes a crucial task. For this purpose, ensemble learning methods have been incorporated extensively, and studies have reported the superiority of these methods over conventional classification methods. This paper provides a comprehensive comparative performance assessment of various ensemble methods for predicting Loan approval in the banking sector. Ensemble algorithms, including bagging, boosting and stacking, are considered with the Neural Network (NN), Decision Tree (DT), Logistic Regression (LR), Naive Bayes (NB), and Support Vector Machine (SVM) as baseline classifiers, which are regarded as benchmarks. The quantitative analysis has been presented in terms of accuracy (ACC), Receiver Operating characteristic Curve (ROC), Area Under the Curve (AUC), Kolmogorov-Smirnov Statistic (KS), Cohen’s Kappa Score (CKS), and Brier Score (BS). The experimental results affirm that ensemble learning performs better than individual learning. LR outperforms other baseline classifiers, whereas the RF (bagging DT) performs the best among the ensemble approaches, followed by XGB and LightGBM, respectively