Symbolic phase transfer entropy method and its application

Abstract

In this paper, we introduce symbolic phase transfer entropy (SPTE) to infer the direction and strength of information flow among systems. The advantages of the proposed method are investigated by simulations on synthetic signals and real-world data. We demonstrate that symbolic phase transfer entropy is a robust and efficient tool to infer the information flow between complex systems. Based on the study of the synthetic data, we find a significant advantage of SPTE is its reduced sensitivity to noise. In addition, SPTE requires less amount of data than symbolic transfer entropy(STE). We analyze the direction and strength of information flow between six stock markets during the period from 2006 to 2016. The results indicate that the information flow among stocks varies over different periods. We also find that the interaction network pattern among stocks undergoes hierarchial reorganization with transition from one period to another. It is shown that the clusters are mainly classified according to period, and then by region. The stocks during the same time period are shown to drop into the same cluster.