Graph Learning Based Financial Market Crash Identification and Prediction

The recognition and anticipation of financial crises play a pivotal role in safeguarding investor interests and maintaining the normalcy of the market order. Therefore, we aim to learn and predict the market abnormal fluctuation from the perspective of its graph structure. In this study, we simplify the financial complex networks with the Planar Maximally Filtered Graph(PMFG) algorithm and analyze their structural characteristic. Then, we integrate the Ollivier-Ricci curvature (ORC) of each network as a manual feature into a graph learning algorithm. We validate the effectiveness of our proposed methodology and engage in a discussion of the future research directions illuminated by our experimental results.