О применении методов машинного обучения для оценки рисков лизинговой компании

Социальные и экономические системы. 2024. № 3(53). С. 152–169.

Чудинова О. С., Ramenskaya A.

 The article examines the practical application of machine learning methods to address the issue of assessing the lessee solvency. The first stage comprises BPMN 2.0 business process modelling for leasing transactions. The second stage envisages the construction of two interpretable models (logistic regression and classification tree) to assess the lessee solvency by means of Python libraries. The models were built taking into account the amount and duration of the leasing agreement, the type of equipment, the legal form of the lessee economic activity and the existence of other contracts. The paper compares the quality of the models and interprets the results. The classification tree model outperforms in terms of quality criteria, whereas logistic regression allows quantifying the impact of the lessee characteristics and predicting the probability of lessee arrears. Based on the findings, the article provides recommendations for the practical application of machine learning models to assess the risk of overdue lease payments. 

Keywords: лизинг, машинное обучение, оценка риска, логистическая регрессия, дерево решений