تجزیه‌ و تحلیل عوامل مؤثر بر ریسک سیستمی بانک‌ها در دوره‌های حباب قیمتی با رویکرد یادگیری ماشین

Purpose: Systemic risk is one of the most significant challenges facing financial systems, particularly banking networks. This type of risk, especially during financial crises and the Bust of price bubbles, can cause widespread disruptions in the financial system's functioning and transmit its negative effects to the entire economy. This research aims to investigate the simultaneous impact of financial variables and banking network structure indicators on systemic risk during the formation and collapse periods of price bubbles. Within this framework, network structure indicators such as centrality measures are analyzed as channels for shock transmission and contagion within the banking network. Alongside, financial variables are considered as factors increasing banks' vulnerability to systemic risks. The combined analysis of these two sets of variables can provide supervisory and regulatory bodies with more efficient analytical and policy tools for managing and controlling systemic risks, while also identifying banks with the highest potential for spreading crises.

Method: Using data from banks listed on the Stock Exchange from 2014 to 2023, this study examines the simultaneous impact of financial variables and banking network structure indicators on systemic risk in the presence of price bubbles, employing the Random Forest algorithm. The results of this model can show the overall importance of each variable in predicting systemic risk. Subsequently, using the Shapley value approach, the role and contribution of each feature in increasing or decreasing systemic risk are analyzed more precisely.

Findings: The research findings indicate that during both the Boom and Bust phases of price bubbles, banks' network structure variables played a more effective role in predicting systemic risk compared to financial variables. However, the relative importance of these two sets of features changes across different stages of the bubble cycle, highlighting the necessity of period-specific analysis. During the bubble formation phase, financial variables, particularly bank size and loan growth, had the highest share in increasing systemic risk. Also, among network structure indicators, betweenness centrality and degree centrality were of high importance. During the bubble collapse phase, the contribution of loan growth increased significantly; this indicates the amplifying role of this variable in triggering systemic risk during crisis periods. Furthermore, the financial leverage variable also gained more importance in risk prediction compared to the previous phase. Conversely, the impact of some network indicators, including closeness centrality and degree centrality, decreased.

Conclusion: These changes suggest that during crisis conditions, the internal characteristics of banks and their financial risk-taking levels play a more significant role in predicting systemic risk than their structural position in the financial network. Particularly during collapse periods, large, highly leveraged banks with an intermediary position in the network have the greatest impact on increasing systemic risk. In contrast, during formation periods, variables such as lending growth and bank size are the most important factors increasing systemic risk, while the role of financial leverage is less prominent in this stage. Based on the research findings, banks, due to differences in financial characteristics and network structure, have varying levels of contribution to systemic risk. Particularly, banks that are larger, have higher leverage ratios, or hold an intermediary position in the banking network play a more prominent role in the transmission of financial crises. Under such conditions, applying uniform prudential regulations to all banks is not only inefficient but could also lead to neglect of high-risk banks and a waste of supervisory resources. Therefore, it is essential for supervisory institutions and policymakers to utilize risk-based supervision frameworks instead of traditional approaches based on uniform requirements; an approach where the intensity of supervision and prudential requirements is determined proportionally to the systemic risk level of each bank. Such an approach not only increases the efficiency of the supervisory system but also enhances its ability to prevent the occurrence of systemic crises and promote financial stability.