Improved Gene Clustering Based on Particle Swarm Optimization, K-Means, and Cluster Matching

Past research has demonstrated that gene expression data can be effectively clustered into a group of centroids using an integration of the particle swarm optimization (PSO) and the K-Means algorithm. It is entitled PSO-based K-Means clustering algorithm (PSO-KM). This paper proposes a novel scheme of cluster matching to improve the PSO-KM for gene expression data. With the proposed scheme prior to the PSO operations, sequence of the clusters' centroids represented in a particle is matched that of the corresponding ones in the best particle with the closest distance. On this basis, not only a particle communicates with the best one in the swarm, but also sequence of the centroids is optimized. Experimental results reflect that the performance of the proposed design is superior in term of the reduction of the clustering error and convergence rate.