A novel fuzzy clustering method based on GA, PSO and Subtractive Clustering
Thanh Le, Lan Vu · 2020
Data clustering is a challenging problem in data science, requiring both accurate determination of the number of clusters and correct clustering of the data. While Fuzzy C-means (FCM) is a powerful algorithm that can cluster data into overlapping groups and converge quickly to local optima, it however depends on the choice of initial parameters and may not always reach a global optimum. In addition, FCM by itself cannot determine the correct number of clusters. In this study, we combine FCM with Genetic Algorithm (GA), Particle Swarm Optimization (PSO) and Subtractive Clustering (SC) algorithms for a novel algorithm, FZGPS, that both determines the correct number of clusters and efficiently constructs these clusters from a given dataset. We show that FZGPS performs effectively on both artificial and experimentally-derived gene expression datasets.