Integration of particle swarm optimization and genetic algorithm for dynamic clustering

Although the algorithms for cluster analysis are continually improving, most clustering algorithms still need to set the number of clusters. Thus, this study proposes a novel dynamic clustering approach based on particle swarm optimization (PSO) and genetic algorithm (GA) (DCPG) algorithm. The proposed DCPG algorithm can automatically cluster data by examining the data without a pre-specified number of clusters. The computational results of four benchmark data sets indicate that the DCPG algorithm has better validity and stability than the dynamic clustering approach based on binary-PSO (DCPSO) and the dynamic clustering approach based on GA (DCGA) algorithms. Furthermore, the DCPG algorithm is applied to cluster the bills of material (BOM) for the Advantech Company in Taiwan. The clustering results can be used to categorize products which share the same materials into clusters.