A Group Search Optimization Method for Data Clustering

Clustering analysis aims to distribute a dataset in-groups in such a way that individuals from the same group have a high degree of similarity among each other, while individuals from different groups have a high degree of dissimilarity among each other. Clustering analysis has become an important mechanism for data exploration and understanding. Evolutionary Algorithms (EAs) have been widely applied for clustering analysis, given their flexibility and capabilities to deal with difficult environments. In this context, Group Search Optimizer (GSO) is a nature-inspired algorithm based on animal searching behaviour and group living theory to solve continuous optimization problems. This paper presents a new evolutionary algorithm for data clustering, named KGSO, which uses a Group Search Optimizer and K-Means approach to perform the clustering task. Experiments were performed on seven benchmark datasets obtained from UCI Machine Learning Repository and seven synthetic datasets to evaluate the performance of proposed algorithm in comparison to other well-known clustering methods from literature.