Using Permutations for Hierarchical Clustering of Time Series

Abstract

Two distances based on permutations are considered to measure the similarity of two time series according to their strength of dependency. The distance measures are used together with different linkages to get hierarchical clustering methods of time series by dependency. We apply these distances to both simulated theoretical and real data series. For simulated time series the distances show good clustering results, both in the case of linear and non-linear dependencies. The effect of the embedding dimension and the linkage method are also analyzed. Finally, several real data series are properly clustered using the proposed method.

Study on Clustering of Free-Floating Bike-Sharing Parking Time Series in Beijing Subway Stations

“…Studies show that the most popular similarity measures in time series clustering are ED and DTW [30]. By the way of shape-based time series, DTW can not only eliminate the matching defect of ED point-to-point and achieve one-to-many matching of data points through bending time to measure time series of unequal length, but also has strong robustness to the deviation and amplitude change of time series [38]. It was proved to be more accurate than the ED [39,40].…”

Time Series Clustering Model based on DTW for Classifying Car Parks

“…For instance, Wang et al [20] evaluated nine distance measures and several variants thereof, and found that ED is the most efficient measure with a reasonably high accuracy while DTW often performs well. DTW is able to overcome the defect of ED point-to-point, and measure time series data with unequal length by warping time to achieve one-to-many matching of data points, which means that it has strong robustness to time deviation and amplitude change [21]. Yao et al [22] realized the initial clustering by DTW and hierarchical clustering, and then put the clustering results into hidden Markov model (HMM) for iterative optimization.…”