سبا طاهري كدخدا

This study aims to predict corporate bankruptcy and financial distress using network features and machine learning models. Two datasets were analyzed in this research. Initially, the data were preprocessed, and the imbalance in the target variable was addressed. Subsequently, five significant features were identified in each dataset, and a network was constructed based on the similarity of companies in these features. In the second dataset, in addition to the similarity method, the correlation method was also used to construct the network. Next, the constructed networks were analyzed to extract valuable information, and companies were clustered into different groups. Several centrality measures were extracted from the network and added to the datasets. Five machine learning algorithms, including Logistic Regression, K-Nearest Neighbors, Support Vector Machine, Decision Tree, and Random Forest, were implemented on the data. The performance of these algorithms was compared in different scenarios using common eva‎luation metrics. The results showed that the network analysis of companies and the addition of network-derived features improved the prediction accuracy of the algorithms. Notably, even in the absence of precise financial information, the network features alone could provide a reasonably accurate prediction of a company's status. Network-based clustering was also very helpful in providing quick predictions. Comparison of network construction methods in the second dataset indicated that the similarity-based method yielded more promising results. Overall, network analysis reveals that corporate networks provide valuable information that can significantly enhance the performance of bankruptcy prediction models and aid analysts and financial managers in making better decisions.

يادگيري ماشين , تحليل شبكه , يادگيري نظارت شده , الگوريتم‌هاي دسته‌بندي , پيش‌بيني ورشكستگي

Machine Learning , Network Analysis , Supervised Learning , Classification Algorithms , Bankruptcy Prediction

Saba Taheri Kadkhoda

Dr. Babak Amiri