An Improved Differential Evolution Algorithm with Opposition-Based Learning for Clustering Problems

Abstract

Differential Evolution (DE) is a popular efficient population-based stochastic optimization technique for solving real-world optimization problems in various domains. In knowledge discovery and data mining, optimization-based pattern recognition has become an important field, and optimization approaches have been exploited to enhance the efficiency and accuracy of classification, clustering and association rule mining. Like other population-based approaches, the performance of DE relies on the positions of initial population which may lead to the situation of stagnation and premature convergence. This paper describes a differential evolution algorithm for solving clustering problems, in which opposition-based learning (OBL) is utilized to create high-quality solutions for initial population, and enhance the performance of clustering. The experimental test has been carried out on some UCI standard datasets that are mostly used for optimization-based clustering. According to the results, the proposed algorithm is more efficient and robust than classical DE based clustering.