Detecting Phase Transitions in Financial Market based on Manifold Learning

In recent years, risks in China’s financial market have increased sharply. It is becoming more important and urgent to understand the potential real state of the market. In this paper, a hybrid manifold learning approach is proposed to detect phase transitions in financial market. By use of Diffusion Maps, the manifold embedding in the financial dynamic system is extracted to identify critical marks of the phase transitions in the financial market. Furthermore, cluster algorithm DBSCAN is employed to discover different periods related to the phase transitions, which would contribute to inform investors about the state of financial markets for investment decisions.