Data Clustering based on Differential Evolution with Modified Mutation Strategy

Abstract

Clustering is an essential and central task of the data mining process. It aims to find meaningful and valuable groups of unlabeled data. Differential Evolution (DE) algorithm has become a successful alternative for clustering because of their ability to achieve good-quality groups of a given dataset. The adopted mutation and crossover strategies can impact the effectiveness of the DE algorithm. This paper proposes a modified mutation strategy for a differential evolution based data clustering algorithm to enhance the cluster solutions’ quality and the convergence speed. The proposed mutation strategy was compared with two normally used mutation strategies such as DE/rand/1 and DE/best/1 and also compared with four nature-inspired metaheuristic algorithms such as Artificial Bee Colony (ABC), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), and Particle Swarm Optimization with Age-Group topology (PSOAG) to study the clustering performance. The experimental tests were conducted on five UCI datasets that are frequently used for metaheuristics based clustering. These tests show that the DE based clustering algorithm with the proposed strategy is more efficient and robust than the others.