Multivariate data analysis using recurrence measures

Abstract

The emergent dynamics of complex systems often arise from the internal dynamical interactions among different elements and hence is to be modeled using multiple variables that represent the different dynamical processes. When such systems are to be studied using observational or measured data, we may benefit from using data from all variables or observations of the system rather than using that from a single variable. In this study, we try to bring out the relative effectiveness of the analysis of data from multiple variables in revealing the underlying dynamical features. For this, we derive the recurrence measures from the multivariate data of standard systems in periodic, chaotic and hyper chaotic states and compare them with that from noisy data. We identify Entropy computed from Recurrence Plot and Characteristic Path Length from recurrence network as the most effective measures that can identify the nature of the dynamical state of the system, and differentiate it from stochastic or noisy behaviour. We find that for different variables, the recurrence measures to be mostly similar for data from periodic states, while they differ for chaotic and hyperchaotic states, indicating that multi-variate analysis is useful for real world systems in the latter states.