Credit risk management and modelling with machine learning algorithms: A model suggestion

The changing landscape of global trade due to globalization and increasing economic integration has led to the credit market becoming a key driver of economic growth. The evolving new world order, characterized by globalization, technological advancements, uncertainty, and heightened competition, has made institutional risks more unpredictable. Consequently, the assessment of credit risk has become more crucial than ever. Effective credit risk management not only enhances companies' competitiveness but also bolsters the stability of the national economy. In the current environment of heightened liquidity needs and increased cost of capital access, trade credit has emerged as a crucial market support mechanism. Therefore, this study aims to utilize customer data and economic indicators, in conjunction with a privately operating company's commercial connections, to create a model that can predict credit risk using machine learning algorithms. The primary objective is to address a gap in the existing literature and provide decision-makers with a tool to assess whether the debt associated with a new customer's order will be paid on time. This approach aims to proactively identify potential losses based on debtors' payment performance and facilitate swift decision-making. Random forest demonstrated the highest classification predictive power and accuracy level among the machine learning algorithms employed. The first six major variables in the model include the average customer overdue period, total payment amount, agreed term for payment, commercial activity period with the customer, the number of days since the last order, and the dollar exchange rate, in that order of importance. Furthermore, it was observed that variables associated with economic data are effective in the model. Further research can focus on developing various credit risk models tailored to specific product types.