Source identification on financial networks with label propagation

Borrowing and lending between banks and firms is the main channel of financial risk propagation, and there have been a number of studies on risk propagation and identification from the financial perspective. Despite complex networks are used as an important analytical tool for risk propagation in the financial system, there are few studies on analyzing financial risk source identification from the perspective of complex networks. With the help of complex network theory, we establish a multi-layer dynamic network between banks and firms, and propose an improved label propagation method for source identification based on the node degree, and this method can be applied to source identification under conditions of incomplete observation. A series of simulation experiments show that the proposed method exhibits a significant advantage in identifying the propagation source of financial risk compared with the original label propagation method. A key conclusion is that when targeting nodes with the highest out-degree, highest in-degree, highest total assets, or highest lent assets, our method encounters significant difficulties in identifying the propagation sources. Conversely, employing an opposite strategy allows us to accurately pinpoint these sources. Moreover, we find that the accuracy of source identification is mainly affected by the proportion of unobserved nodes, while the number of sources and the average connectivity of the network have relatively little effect. This study provides a new perspective for the study of risk propagation identification in financial network systems.