Multiple temporal pattern detection and predictability analysis of complex time-evolving systems

This paper presents a new method, called multiple temporal pattern recognition (MTPR), that is capable of detecting multiple temporal patterns for characterizing and predicting events of interest in the time-evolving system data. The MTPR method embeds time series data into multiple phase spaces with various dimensions and time delays. Then it clusters the embedded data to detect the preliminary temporal patterns. We further performed a three-stage statistical predictability analysis to evaluate the confidence of the detected temporal patterns. At the first stage, we introduced a new predictability measure, p_m, to evaluate the temporal patterns and then apply the statistical logistical regression to further validate these patterns. Experimental results are also included to illustrate the MTPR method.