Clustering Algorithm Based On Differential Evolution Algorithm

With the advance of the technology of science, especially the development of theinformation industry, the human society has stepped into a new period of information. Facedwith the growing flood of information, the work which is how to extract useful information, andhow to have the necessary collation and analysis of information, has currently become one of theissues which have most research and concern. Clustering analysis is an important method todivide the data into meaningful groups. Its result is an important guidance, so it is used widelyincluding: psychology and other aspects of social sciences, biology, statistics, pattern recognition,information retrieval, machine learning and so on.The classical clustering analysis methods include partition-based clustering algorithm,hierarchical clustering algorithm, density-based clustering algorithm, grid-based clusteringalgorithm and model-based clustering algorithm and so on. This article focuses on the two mostwidely used clustering algorithms: partition-based clustering algorithm and hierarchicalclustering algorithm. For the partition-based clustering algorithm it focuses on the K-Meansclustering algorithm and K-Modes clustering algorithm; for the hierarchical clustering algorithmit focuses on the Ward clustering algorithm. But these clustering algorithms are sensitive to thechoice of initial cluster centers or shortcoming that plunges into a local optimum prematurely.Taking the optimization algorithms into clustering algorithms is an effective method tosolve the above shortcomings of clustering algorithms. The famous optimization algorithmsinclude: Differential Evolution algorithm, Genetic algorithm, Ant Colony Optimization,Particle Swarm Optimization, Artificial Fish-swarm algorithm and so on. These algorithms havedifferent characteristics, widely used in solving various optimization problems. This articleparticularly introduces the Differential Evolution algorithm. It is a simple structure, easy to use,fast and robust features. And it is ideal for solving various numerical optimization problems. Soit is a most valuable optimization algorithm.This article mainly takes the Differential Evolution algorithm, into the K-Means clusteringalgorithm, K-Modes clustering algorithm and Ward clustering algorithm for optimization. It cansolve the problem of optimizing the attributes center effectively. Let the three cluster clustering algorithms above have better results. With taking the Differential Evolution algorithm into theK-Modes clustering algorithms, it also improves the K-Modes clustering algorithm based on theDifferential Evolution algorithm by retaining the better clustering offspring. The experimentsprove that the improved clustering algorithm is better. When taking the Differential Evolutionalgorithm into the K-Means clustering algorithm, it has better results with the respectivecomparison of the K-Means clustering algorithm based on the Genetic algorithm and theEvolutionary Programming. When taking the Differential Evolution algorithm into the Wardclustering algorithm, it also has better results with the comparison of the Ward clusteringalgorithm based on the Genetic algorithm.