Particle Swarm Optimizer for High-Dimensional Data Clustering

This chapter aims at developing effective particle swarm optimization (PSO) for two problems commonly encountered in studies related to high-dimensional data clustering, namely the variable weighting problem in soft projected clustering with known the number of clusters k and the problem of automatically determining the number of clusters k. Each problem is formulated to minimize a nonlinear continuous objective function subjected to bound constraints. Special treatments of encoding schemes and search strategies are also proposed to tailor PSO for these two problems. Experimental results on both synthetic and real high-dimensional data show that these two proposed algorithms greatly improve cluster quality. In addition, the results of the new algorithms are much less dependent on the initial cluster centroids. Experimental results indicate that the promising potential pertaining to PSO applicability to clustering high-dimensional data.