Complexity measure by ordinal matrix growth modeling

We present a new approach based on the modeling of the behavior of the number of ordinal matrices derived from time series, as a function of the embedding dimension. We show that the number of distinct ordinal matrices can be used for determining whether the dynamics are regular or chaotic by means of the periodicity (\(\mu \)), quasiperiodicity (\(\alpha \)) and nonregularity (\(\lambda \)) index herein defined. We verify that \(\lambda \) behaves similarly to the Lyapunov exponent and therefore can be used for measuring complexity in time series whose underlying equations are unknown. Moreover, the combination of \(\mu \), \(\alpha \) and \(\lambda \) enables us to distinguish between deterministic and stochastic data. We thus propose the variation law of the number of ordinal matrices characterizing the random walk.