Dynamic programming-based optimization for segmentation and clustering of hydrometeorological time series

Abstract

In this study, we propose a new segmentation algorithm to partition univariate and multivariate time series, where fuzzy clustering is realized for the segments formed in this way. The clustering algorithm involves a new objective function, which incorporates an extra variable related to segmentation, while dynamic time warping (DTW) is applied to determine distances between nonequal-length series. As optimizing the introduced objective function is a challenging task, we put forward an effective approach using dynamic programming (DP) algorithm. When calculating the DTW distance, a DP-based method is developed to reduce the computational complexity. In a series of experiments, both synthetic and real-world time series are used to evaluate the performance of the proposed algorithm. The results demonstrate higher effectiveness and advantages of the constructed algorithm when compared with the existing segmentation approaches.

Temporally Aligned Segmentation and Clustering (TASC): A generalized framework for behavior time series analysis

“…shown to aid time series segmentation and clustering, including in regard to HAS [16][17][18] . In this study, we utilize iterative temporal alignment to provide a novel approach to automated behavior segmentation and clustering which provides motif boundary refinement robust to temporal variability.…”

Temporally aligned segmentation and clustering (TASC) framework for behavior time series analysis

“…DBA iteratively refines an average sequence to minimize its distance to a cluster of time series. These methods have also been previously shown to aid time series segmentation and clustering in tasks such as HAS 26 – 28 . In this study, we utilize iterative temporal alignment to automate behavior segmentation and clustering to provide motif boundary refinement robust to temporal variability.…”