Composite similarity measure algorithm
Yan Wang, An Yunjie · 2016
Representation of time series and similarity measure are the basis of time series data mining, but the common method of similarity measure isnt considering the morphology features, or isnt considering it at all, or is considering one of morphology and statistical feature. Aiming at this problem, proposing the composite representation of Symbolic Aggregate approximation (SAX) and bending value of angular point that represents time series. SAX, as the most commonly representation method of statistical feature, can accurately reflect the average of sequences, and the bending value of angular point, that is a better robustness method, can precisely represent the trend change of sequences. It can mirror the general information of sequences and make sequence simply by combining the two methods. At the same time, applying the composite distance algorithm has highly quality to measuring the similarity, that can more effectively show the difference in the between of sequences and meet the demand of similarity measure. Experimental results show that is not only simply and highly accuracy and better robustness, but also obtain good result of the similarity measure.