Spatio-temporal data evolutionary clustering based on MOEA/D

Jingjing Ma, Yanhui Wang, Maoguo Gong, Licheng Jiao, Qingfu Zhang

Abstract: Clustering the data evolve with time, which is termed evolutionary clustering, is an emerging and important research area in recent literature of data mining, and it is very effective to cluster the dynamic data. It needs to consider two conflicting criteria. One is the snapshot quality function; the other is the history cost function. Most state-of-the-art methods combine these two objectives into one and apply a single objective optimization method for optimizing it. In this paper, we propose a new evolutionary clustering approach by using a multi-objective evolutionary algorithm based on decomposition (MOEA/D) to optimize these two conflicting functions in evolutionary k-means algorithm (EKM). The experimental results demonstrate that our algorithm significantly outperforms EKM.