Early Identification of Debt Collection Cases with Low Recovery Potential Using Machine Learning

Abstract

Machine learning methods offer new opportunities for improving managerial decision-making and economic efficiency in the management of mass receivables portfolios. This article assesses their potential application in the early identification of debt collection cases with low recovery potential. From the perspective of management and economics, the early recognition of such cases may support more rational case prioritisation, better allocation of operational resources, reduction of ineffective collection costs, and improved efficiency of debt collection processes. The empirical study was based on a large real-world dataset provided by a debt collection entity. The target variable was defined as information on whether the total amount of payments was lower than the purchase price of the receivable. Three classification models were compared: logistic regression, Random Forest, and XGBoost. The best predictive performance was achieved by the XGBoost model. The results indicate that tree-based models identify low-recovery-potential cases more effectively than the linear model. In addition, SHAP analysis was applied to increase model interpretability and to link predictive results with expert and managerial knowledge. The article shows that machine learning may support portfolio segmentation, case prioritisation, and resource allocation in debt collection management. However, such methods should be treated as decision-support tools rather than instruments for fully automating debt collection decisions.