A New Ensemble Learning Method for Temporal Pattern Identification

Abstract

In this paper we present a method for identification of temporal patterns predictive of significant events in a dynamic data system. A new hybrid model using Reconstructed Phase Space (MRPS) and Hidden Markov Model (HMM) is applied to identify temporal patterns. This method constructs phase space embedding by using individual embedding of each variable sequences. We also employ Hidden Markov Models (HMM) to the multivariate sequence data to categorize multi-dimensional data into three states, e.g. normal, patterns and events. A support vector machine optimization method is used to search an optimal classifier to identify temporal patterns that are predictive of future events. We performed two experimental applications using chaotic time series and natural gas usage series related to the natural gas usage forecasting problem. Experiments show that the new method significantly outperforms the original RPS framework and neural network method.