A New Representation and Similarity Measure of Time Series on Data Mining

Abstract

Representation and similarity measure of time series is the research basic of the time-series data mining. This paper uses ESAX (extended symbolic aggregate approximation) representing the time series similarly and raises an improved time series method of similarity measure ESSVS (ESAX statistical vector space) based on the statistics symbolic vector space method. ESSVS measure the time series similarity by the eigenvector, which comes from the statistical features of symbolic sequences. The experiment results show that the similarity measure method is simple and effective, with a better clustering performance. Compared to the classic method of similarity measure, this new method could improve the measure accuracy with a smaller complexity, and it can be applied to different areas of time series.