Financial crisis prediction is essential in preventing financial problems as its monitoring indicators help regulators judge the probability of future crises. In this context, the activities of the scientific community have been focused on the dynamics of single/multiple sequences and utilized unsupervised/supervised methods for financial crisis prediction. It is noteworthy that the cross-correlation between the risks of multiple economic entities makes financial network analysis paramount in crisis prediction. Focusing on this point, we propose a multilayer supervised network analysis (MSNA) method to train the multilayer network, and select the most suitable layer for financial crisis prediction. Specifically, we use 37 crucial stock market indices from 4 continents to create successive multilayer financial networks with 120-day windows and 1-day step by Pearson cross-correlation (PCC), variance decompositions (VD), transfer entropy (TE), minimum spanning tree (MST), directed MST (DMST), planar maximally filtered graph (PMFG) and directed PMFG (DPMFG) methods. Based on the multilayer network, we embed the graph neural network classification (GNNC) model and train the dynamic multilayer networks at each window scale (240,120, and 60 days). Finally, we conclude that the accuracy of the short window (60 days) is significantly higher than that of the long window. The network constructed by PCC with MST is the most suitable for short sequence (60 days) crisis prediction (AUC = 0.959), and the network constructed by TE with DMST is the most suitable for long sequence (240 days) crisis prediction (AUC = 0.772).

The systemic risk of banks is a multi-dimensional factor correlation and multi-interbank contagion, forming a multi-dimensional multi-correlation (MDMC) contagion risk. After superposition, the multiplier effect magnifies the destructive power of a single impact. In this study, the weighted average method is used to integrate the four interbank contagion paths of jump-diffusion, interbank lending, stock price information, and common assets, establish the multi-level interbank contagion matrix, apply the quantile estimation of the spatial Dubin panel model, estimate the MDMC infectious bank conditional risk value, and decompose and identify the systemically important risk factors, systemically important banks, and systemically vulnerable banks. The following conclusions were drawn. First, the superposition contagion effect of MDMC networks is significant. Second, the systemic importance of default risk, interest rate risk, liquidity risk, and the GDP growth rate of the banking industry is high, followed by changes in stock market returns and investor sentiment. Third, the four major state-owned banks have an MDMC network contagion effect, which has the characteristics of systemic importance and vulnerability.

The major issue in the evolution of housing prices is risk of housing price contagion. To model this issue, we constructed housing multilayer networks using transfer entropy, generalized variance decomposition, directed minimum spanning trees, and directed planar maximally filtered graph methods, as well as China’s comprehensive indices of housing price and urban real housing prices from 2012 to 2021. The results of our housing multilayer networks show that the topological indices (degree, PageRank, eigenvector, etc.) of new first-tier cities (Tianjin, Qingdao, and Shenyang) rank higher than those of conventional first-tier cities (Beijing, Shanghai, Guangzhou, and Shenzheng).

银行风险传导研究是系统性金融风险测度与防范的重点. 文献中主要研究静态银行网络拓扑结构和银行系统性风险量化, 较少考虑银行网络之间的状态转变. 针对上述问题, 提出时变状态网络模型. 根据模型, 首先用kmeans方法对各个时间段的银行网络分类, 然后通过有向最小生成树(DMST, directed minimum spanning tree)分析每一类银行网络的拓扑结构, 最后联合利用平面最大过滤图(PMFG, planar maximally filtered graph)方法构建时变银行状态网络, 该网络可用作银行风险源头的寻找和传导时变性分析. 利用时变状态网络模型研究我国15家上市商业银行2007年第四季度到2019第一季度同业拆借数据. 结果表明, 银行状态网络之间的短期连续跳跃性可有效描述金融危机的发生, 比如2008年全球金融危机发生前后出现了两个状态间的短期跳跃, 从2013年“钱荒”到2015年的股灾阶段, 先后出现了四个状态间的短期跳跃. 同时, 各个有向银行状态网络的出度和传染效应成正比, 入度和银行面临风险的稳健程度成反比. 时序的银行状态网络具有记忆性特征, 这可以为央行防范系统性风险提供决策依据.