Adaptive G–G clustering for fuzzy segmentation of multivariate time series

Abstract

In this paper, Gath–Geva (G–G) fuzzy clustering is extended to adaptively segment hydrometeorological multivariate time series. First, KPCA is used to extract principle components of multivariate time series to remove the impacts of redundant and irrelevant variables. Then, taking the time information into account, the segmentation of principle components of multivariate time series is derived with the modified Davies–Bouldin Index and adaptive G–G fuzzy clustering. In the experiment, our proposed algorithm is applied on the real-world hydrometeorological multivariate time series collected every 6 min with length $$N=720$$ . Comparison with the existing segmentation algorithms, our proposed algorithm proves the applicability and usefulness in hydrometeorological multivariate time series analysis.