Machine Learning Based Risk Management of Credit Sales in Small and Mid-Size Business

Abstract

This is a study that uses ML algorithms applications for effective credit risk prediction and management in small and mid-size businesses (SMBs). One of the ways this was achieved was by using comprehensive data sets, which consisted of historical credit sales transactions, customer demographics, and economic indicators. As a result, four specific ML algorithms, namely logistic regression, decision trees, random forest and gradient boosting, were assessed as the methodology. Findings show that gradient boosting yielded the best results, reaching an accuracy score of 90 %, precision of 89 %, recall value of 91 %, F1-score of 90 %, and area under the receiver operating characteristic curve is 0.95. Logistic regression has shown highly competitive results, in excess of 85% accuracy, and an AUC-ROC of 0.91. The findings demonstrate that credit history, the income level, and the age of the client are the most critical features in credit risk analysis of the SMBs.