Hierarchical Algorithm to Match Similar Time Series Pattern

A time series is first decomposed into a trend part and some detail parts via empirical mode decomposition. Then the trend part is represented in the form of piecewise linear segments to reduce its dimensionality and these segments are transformed further into a 0-1 string to fit the fast matching algorithm. After clustering the transformed trend series, rough similar time series will be obtained. Finally by calculating the distance of the clustered series, accurate similar series patterns are reached. Experiments show that this hierarchical approach is effective.