Clustering Analysis Based On Intelligent Optimization Algorithm And Applications

In recent years, with the advance of the technology of science and the prevalence of the applications of database, data mining is becoming one of the hottest research topics in the field of the information decision-making in the world, meanwhile, and is a new subject formed with the development of information technology. As an effective analysis method of data mining, cluster analysis has become a hot research project in the field of data mining.There are many kinds of clustering analysis method, in this paper, mainly introduces C-Means and fuzzy C-Means partition-based clustering methods. The two algorithms plunge into a local optimum prematurely because they are sensitive to the choice of initial cluster centers and shortcomings. so, the intelligent optimization algorithm was presented.Introducing intelligent optimization algorithms into clustering analysis is an effective solution to overcoming the shortcomings of the traditional clustering algorithms. Intelligent optimization algorithms mainly covers the following:differential evolution, ant colony algorithm, artificial immune algorithm, clone selection algorithm, genetic algorithm, simulated annealing algorithm and other natural inspired methods. All of these algorithms have their characters and have been applied to many applications. This paper focuses on realizing the hybrid clustering algorithms based on clone selection algorithm, genetic algorithm and simulated annealing algorithm, and on applying them into designing codebook in the vector quantity. The main research jobs are given as follows:Firstly, to solve such problems as too many parameters and low clustering accuracy in simple clone selection algorithm, a new local learning operator based on artificial-immune multi-objective clustering algorithm is proposed. In the algorithm, a new local learning operator is proposed to overcome the shortage of single clustering index. Two different indexes are proposed to optimize simultaneous and a multi-objective clustering algorithms. The proposed clustering is applied to an artificial data set and UCI data set and the simulation results show that the algorithm has high clustering accuracy. Secondly, to solve premature phenomenon and falling into local optimum of genetic algorithm, the simulated annealing algorithm is fused to the genetic algorithm and a simulated annealing is presented based on genetic clustering algorithm, a new effective SA, crossover operator and mutation operator proposed for fitting the partition-based chromosome coding. In addition, the Euclidean distance is replaced by the kernel function distance to improve the performance of the proposed algorithm further. We also applied the proposed algorithm to vector quality. The experimental results show that the algorithm can achieve good results in most data sets and the robustness of algorithm is also very good.Finally, since LBG algorithm is very sensitive to initial codebook, clone selection algorithm is introduced and splitting method is adopted to generate the initial codebook. Besides, the similarity metric based on Euclidean distance can only reflect local consistency of the clustering instead of the clustering global consistency. Clone selection clustering method in terms of manifold distance is put forward.Besides, we also applied the proposed algorithm to image compression. It is proved through the experiment that this algorithm is characterized by better performance compared with other algorithms.