A Comparative Analysis of Macroeconomic Indicators in Optimising Credit Risk Prediction

Credit risk modelling plays a role in financial institutions' evaluation of the creditworthiness of borrowers and in managing lending risks effectively. Feature selection is critical in developing robust and interpretable credit risk models. This paper presents a comparative analysis of various macroeconomic indicators applied to linear regression analysis for credit risk modelling. Specifically, this study compares three feature selection techniques: clustering feature, feature combination and correlation. This comparative study aims to identify the most compelling feature selection strategy regarding model performance, interpretability, and computational efficiency. This study employs a real-world dataset comprising various macroeconomic indicators in Indonesia and a financial institute's default scores to train and evaluate the linear regression models. Experimental results demonstrate that the clustering feature outperforms feature combinations and correlation features in optimizing credit risk modelling by achieving higher predictive accuracy with fewer features and improved interpretability.