A New Time Series Outlier Pattern Detection Approach based on Markov Model

Time series pattern outlier represents a pattern with abnormal behavior that is significantly different from other patterns in the time series and can induce a bias in the decision-making process related to design, operation, and management. This research focuses on the problem of detecting anomalous patterns from time series data. The main consideration is the fact that normal pattern can be modeled by some statistical models and anomalous patterns cannot. Combined with characteristics of time series data, this research develops an effective and accurate outlier pattern detection approach SAX_ePST to detect anomalous patterns in time series. Experiments indicate the proposed methods are fast and correctly identifying anomalous pattern.