A Population-Based Clustering Technique Using Particle Swarm Optimization and K-Means

Abstract

Abstract. Population-based clustering techniques, which attempt to integrate particle swarm optimizers (PSOs) with K-Means, have been proposed in the literature. However, the performance of these hybrid clustering methods have not been extensively analyzed and compared with other competitive clustering algorithms. In the paper, five existing PSOs, which have shown promising performance for continuous function optimization, are hybridized separately with K-Means, leading to five PSO-KM-based clustering methods. Numeric experiments on nine real-life datasets show that, in the context of numeric data clustering, there exist no significant performance differences among these PSOs, though they often show significantly different search abilities in the context of numeric function optimization. These PSO-KM-based clustering techniques obtain better and more stable solutions than some individual-based counterparts, but at the cost of higher time complexity. To alleviate the above issue, some potential improvements are empirically discussed.

Flow-based Anomaly Detection Using Access Behavior Profiling and Time-sequenced Relation Mining

“…Particle swarm optimization(PSO) [6] is a computational method that optimizes a problem by iteratively trying to improve a candidate solution with regard to a given measure of quality. It is broadly used in optimizing classification [27] and clustering approaches [28,29]. In this section ,we propose a novel linear grouping method PSOLGA with a combination of PSO and LGA, which is able to optimize the resampling process in LGA and output more stable grouping result.…”