Using Improved Glowworm Swarm Optimization Algorithm for Clustering Analysis

Abstract

High-quality clustering algorithms play an essential role in the data analysis. Traditional clustering algorithms are susceptible to initial cluster centers, which leads to the degradation of clustering quality. Clustering analysis based on Swarm Intelligence (SI) optimization algorithms can solve the problem of traditional clustering algorithms by the better ability to find the optimal solutions, such as Glowworm Swarm Optimization (GSO) algorithm. The GSO-based clustering analysis can use the multi-modal optimization ability to search for the optimal cluster centers. This paper proposes two clustering techniques based on the GSO algorithm: one is to realize self-organizing clustering of data through improved GSO algorithm, and the other is to hybrid the improved GSO self-organizing clustering algorithm with the k-means algorithm. The proposed algorithms test on the Iris data set. The experimental results show that the enhanced GSO algorithm proposed in this paper can efficiently realize self-organization clustering of data without initializing cluster centers and cluster number, and the proposed hybrid algorithm has better clustering results over four clustering algorithms.