Particle swarm optimization algorithm with environmental factors for clustering analysis

In view of the model of bird flocking, particle swarm optimization (PSO) is a promising method to tackle optimization. This study benefits from the fact that the living environment affects behaviors of the bird flocking. That is, a well-defined environmental factor can improve the performance of PSO. Thus, the environment factor is taken into account to inspire the robust behaviors of bird flocking in depth. Specifically, it not only can carry out effective searching in limited flying space, but also can strengthen the social behaviors of individual. In the field of clustering, it can be regarded as a search optimization issue. Like the utilization of some useful information generated in its process, environment factor is considered and environment factor-inspired PSO (EPSO) is proposed in this study. To take full advantage of EPSO for solving issue of clustering, we divide its process into two stages. In the first stage, the environment factor is imported as a refined search technology to achieve the multi-local optimums with high probability. In the second stage, the manifold information, i.e., individual, swarm and environment factors, is considered to improve its global search capacity. Such an approach can effectively overcome the defect of PSO being prone to being trapped in local optima. To demonstrate the validity of our approach, EPSO, conventional PSO, genetic algorithm, $$K$$K-means, artificial bee colony and hybrid ABC are compared with benchmark document collections. The experiment results indicate that EPSO performs better than these state-of-the-art clustering algorithms in most cases.