金融システムにおける大変動の検出に関する理論解析

摘要

We have tested the methods to detect early warning signals for financial systems. We have tried to detect early warning signals using the idea of the dynamical network marker, which can detect the qualitative change of nonlinear dynamics. We have proposed two methods to detect the dominant group. We have compared the above method with the method based on the Koopman mode analysis and validated the detection abilities of the dynamical network marker. Also, we have proposed two methods that overcome the drawback of KMA. The truncated KMA method increases detectability of change points by considering the contribution of each eigenvalues. The shift DMD method can be used for a more general system because it can be applied to models including constant terms. Even if we cannot detect the precursors of such systemic catastrophes, detecting the systemic catastrophes soon after they occur is meaningful. We have clarified that combining multiple methods can improve detection capability.