A Comparative Study of Machine Learning Algorithms for Enhanced Credit Default Prediction

Abstract

In today’s competitive financial arena, accurate credit default prediction is very important for sustaining the stability and profitability of banks. This research study presents a comparative analysis of various machine learning algorithms, which are used for forecasting the likelihood of credit default. Six diverse algorithms—Support Vector Machine (SVM), K-Nearest Neighbor (K-NN), Logistic Regression, Decision Tree (DT), Gaussian Naive Bayes, and Random Forest (RF)—are used to construct the predictive comparison. All the models were trained and evaluated by using an investment dataset obtained from a private bank located in Dhaka, Bangladesh. The results of the study indicate that the Random Forest (RF) and Decision Tree (DT) models have achieved higher accuracy in predicting the outcomes when compared to other machine learning methods, with an accuracy of 92 and 94%, respectively. This study also highlights the importance of feature selection and prediction boosting in order to optimize the credit default prediction rates.