Cluster Analysis Based On Particle Swarm Optimization Algorithm

The characteristics of the clustering problem are that we don’t know the category and other prior knowledge in a batch of sample in advance. The only classification is based on the characteristics of the sample. Clustering is often the first step in data mining. In today’s society, with a large amount of data accumulating over a long period of time, our life becomes overwhelmed with huge amounts of data. It is urgent and necessary to analysis and dig up the effective information for human beings’ needs from a large amount of data. For example, in the electronic commerce, the human need to know which kind of products has the highest satisfaction for consumers. First, we can assign the entire consumer groups in different classes randomly. Then, we carry out cluster analysis by using clustering algorithm according to certain rules and characteristics. We can achieve better promotion effect and commercial profits by using the results of clustering and promoting for different kinds of customer requirements. This paper introduces the research background, principle and algorithm flow description of particle swarm optimization and clustering analysis,also introduces the application of PSO in function optimization and data clustering,and programs to realize the particle swarm optimization(PSO) algorithm application in the above two aspects respectively in MATLABR2013 a programming environment.Nature is the source of all kinds of human thoughts, engineering principle and invention. In the process of learning from nature, human society continues to get inspiration, and promote the development and progress of the philosophy, medicine,military science, computer science, information management and control, etc.Intelligent computing is the result of human exploring and learning from living nature constantly. By imitating the collective wisdom of biological community in foraging,danger signals, we study the mechanism of biological group’s social division of labor and cooperation. And there are a variety of intelligent optimization algorithms produced simultaneously, such as particle swarm optimization algorithm, ant colony algorithm, leapfrog algorithm, simulated annealing algorithm, bacterial foragingalgorithm and swarm algorithm, etc. In nature, in the face of a piece of food in the natural space, birds will search the food collaboratively, each bird will learn from the surrounding birds, and constantly adjust their position and speed information.Through their own optimization and global optimization, the flock can improve the efficiency of the searching for food. Particle swarm optimization algorithm is a model to imitate birds foraging behavior together. Through the long-term observation of birds foraging phenomenon, human beings continue to search for rules, and do the mathematical modeling, algorithm analysis design and computer programming for foraging behavior. This algorithm is applied to optimization of function originally,then applied to all areas of life gradually, such as data mining, neural network training,mechanical fault diagnosis, electricity, transportation department, and so on, through the development and improvement of particle swarm optimization.This paper introduces the basic theory and the classical algorithm of clustering analysis, and emphatically analyzes two clustering methods: the K- means clustering and K-medoids clustering. We must recognize that clustering analysis is a kind of exploratory analysis. Using different clustering algorithm, the conclusion is often different. This phenomenon has also encouraged us to find more satisfactory clustering algorithm. First of all, the paper presents the working principle and the algorithm flow of the particle swarm clustering algorithm, and uses the computer programming to achieve the goal of data clustering. Then the article compares the particle swarm optimization algorithm and K- means clustering algorithm respectively:particle swarm clustering algorithm is easy to fall into local extreme points. And its search accuracy is not high. K-means clustering results are greatly influenced by the initial clustering center, etc. Finally according to the disadvantages of the two algorithms, the paper presents an improved particle clustering algorithm based on K-means. In this paper, the principle and the corresponding algorithm are described.Through the computer programming to achieve the same data clustering, and comparison with the particle clustering algorithm clustering effect, we can clearly discover that the improved particle swarm optimization clustering is better, which can finish the data clustering in less number of iterations. At the same time, the paperanalyzes the research status of particle swarm optimization algorithm and clustering algorithm and puts forward the deficiencies of them. Especially, the particle swarm algorithm appears late relatively. The foundation of its basic theory is weak. The particle swarm algorithm is lack of mathematical deduction analysis in the convergence of this algorithm and its parameter settings. The part of outlook puts forward new problems and finds the direction of the efforts in future.