Multi-Objective Data Clustering using Variable-Length Real Jumping Genes Genetic Algorithm and Local Search Method

In this paper, we present a novel multi-objective evolutionary clustering approach using variable-length real jumping genes genetic algorithms (VRJGGA). The proposed algorithm that extends jumping genes genetic algorithm (JGGA) [1] evolves clustering solutions using multiple clustering criteria, without a-priori knowledge of the actual number of clusters. Some local search methods such as probabilistic cluster merging and splitting are introduced in VRJGGA for the clustering improvement. Experimental results based on several artificial and real-world data show that VRJGGA can obtain non-dominated and near-optimal clustering solutions in terms of different cluster quality measures and classification performance.