A modified algorithm for clustering based on particle swarm optimization and K-means

Clustering is a technique that can divide data objects into meaningful groups. Particle swarm optimization is an evolutionary computation technique developed through a simulation of simplified social models. K-means is one of the popular unsupervised learning clustering algorithms. After analyzing particle swarm optimization and K-means algorithm, a new hybrid algorithm based on both algorithms is proposed. In the new algorithm, the next solution of the Problem is generated by the better one of PSO and K-means but not PSO itself. It can make full use of the advantages of both algorithms, and can avoid shortcomings of both algorithms. The experimental results show the effectiveness of the new algorithm. First reduces the dataset´s dimensionality using the Singular Value Decomposition (SVD) method, and only then employs various clustering techniques. Besides its simplicity, and its ability to perform well on high dimensional data, it provides visualization tools for evaluating the results. It was tested on a variety of datasets, from classical benchmarks to large-scale gene-expression experiments. It is configurable and expendable to newly added algorithms.