Cluster analysis is one of the important techniques in data mining and exploratory data analysis, and it is effective in discerning the structure of massive amounts of data and unraveling its complex relationship. In this paper, an improved hybrid algorithm which combines linear particle swarm optimization with a modified chaotic strategy is proposed. On the one hand, the time-varying acceleration coefficients and inertia weights can balance the exploration and exploitation. On the other hand, the modified chaotic strategy can increase the population diversity for avoiding encountering stagnation. Modified chaotic linear particle swarm optimization (MCLPSO) searches through data set for given cluster centers and can efficiently find better solutions by generating new particles. Comparing with other four methods on six datasets for clustering problem, the results show that the new algorithm has better clustering performance, and it can improve the quality and stability of clustering results in a certain extent.

data clustering, particle swarm optimization, modified chaotic strategy.