Trigger A Complex Network Of Time Series Analysis Methods And Applications, Based On The Granger Causality

This paper proposes an approach of integrating the Granger causality and com-plex networks into characterizing time series with noise perturbations in both the time and the frequency domains. Representative measurements describing the physical and topological properties of a weighted and directional complex network are used in the characterization. Strikingly, the proposed approach shows a strong competency in dis-tinguishing typical dynamical behaviors of time series with noise perturbations. Partic-ularly, this approach, when applied to the human electrocardiograms from the MIT-BIH datasets, becomes practically useful in discriminating the normal sinus rhythm subjects from those with arrhythmia symptom. In addition, the research of nonlinear Granger causality has just started, this paper also innovatively integrates artificial neural net-works into Granger causality analysis and builds up the theory framework systemati-cally, designing a new measurement method for nonlinear Granger causality.