An advanced segmental semi-Markov model based online series pattern detection

The online pattern detection technology is an important part of the time series analysis, and some methods have been proposed, in which subsequence matching based window-sliding is popularly applied. For window-sliding, Euclidean distance and dynamic time warping (DTW) are always used as subsequence matching, but they have the drawbacks of sensitivity and expensive computational load respectively. Recently, the model based method is introduced into the field of online pattern detection, especially, the segmental semi-Markov model shows better performance than sliding methods in many aspects. However, it has some limitations, e.g., it is difficult to estimate the parameters of the model, and nowaday methods are too rough, etc. In this paper, the advanced segmental semi-Markov model is proposed to improve the existed segmental semi-Markov model and it is successfully demonstrated on real data sets, including financial and medical data.