A Hybrid Evolutionary Approach for Multi-Objective Clustering

Description

This research proposes a novel multi-objective evolutionary clustering approach using the Hybrid Variable-length Real Jumping Genes Genetic Algorithm (H-VRJGGA). The proposed algorithm that extends the Real-coding Jumping Genes Genetic Algorithm (RJGGA) evolves clustering solutions using multiple clustering criteria, without a-priori knowledge of the actual number of clusters. Traditional clustering algorithms have been limited to a single clustering objective and often fail to detect meaningful clusters because most practical data sets are characterized by a high-dimensional, inherently sparse, data space. Inspired by the inherent multi-objective nature of data clustering, previously the investigators presented several evolutionary approaches to data clustering using multiple objectives. Their results showed that previous approaches were capable of producing diverse clustering solutions while maintaining the consistency and convergence of the non-dominated solutions. However, in some rare cases, the solutions may be too diverse. This is due to the additional diversity, which is naturally provided by the jumping operations introduced in RJGGA. In this project, they extend this idea by introducing a hybrid application of jumping gene operation for multi-objective clustering, which alleviates the above difficulty. The proposed approach also includes some local search strategies for clustering improvement.