كوثر فرج پورمژدهي

Small and medium-sized enterprises (SMEs) play a significant role in the economic development of societies. The presence of these companies, both in developed and developing countries, promotes entrepreneurship, innovation, technological change, industry dynamism, job creation, and income generation. Despite these features, one of the serious and undeniable limitations for the survival, growth, and development of SMEs is the lack of capital and access to resources. One suitable solution is leveraging the benefits of supply chain financing. Therefore, eva‎luating the credit risk of SMEs that benefit from this type of financing is crucial. In this regard, this study investigates the combination of graph neural network models with the Ball-Mapper algorithm, one of the topological data analysis methods, and ultimately presents the BallMapper- Network (BM-GNN) model. The results show that using the Ball-Mapper algorithm for feature selection enhances the efficiency and positive impact of graph neural network models. Additionally, integrating network features significantly improves the accuracy of the eva‎luation. The accuracy of the proposed model reaches 93.56% when the network features derived from the selected indicators by the Ball-Mapper model are combined with all financial indicators of the companies. This research offers a significant advancement in credit risk prediction and highlights the potential capabilities of graph neural networks and topological data analysis in improving financial decision-making processes.

ارزيابي ريسك اعتباري , بنگاه‌هاي كوچك و متوسط , تأمين مالي زنجيره تأمين , تجزيه و تحليل توپولوژيكي داده‌ها , شبكه عصبي گراف

Credit Risk Assessment , Small and Medium-sized Enterprises , Supply Chain Finance , Graph Neural Network , Topological Data Analysis

Kosar Farajpour Mojdehi

Dr. Babak Amiri