Theoretical Analysis on Detection of Major Change in Financial Systems

Abstract

Since crashes of financial bubbles cause damage to our society, it is important to predict the crashes and take necessary actions. The dynamical network marker can be applied to such real-time precursor detection in multivariate time series data of financial systems. If we do not know the mathematical model of the time series data, we have to choose the dominant group heuristically. We propose two methods to choose the dominant group. We compare the above method with the other methods based on the Koopman mode analysis （KMA）and we propose two methods that overcome the drawback of KMA. We test these methods in stock market data.