Uncertain Cluster Analysis Based On Particle Swarm Optimization

With the continuous development and improvement of communication and Internet technology, we have entered a highly information-intensive era of big data, the financial, Internet, military, geographical and other related sectors have produced countless data, it may be hidden many valuable information behind of these data, how to use data mining techniques to deal with these complex data has become a research focus of the moment.Cluster analysis is one of the most dynamic areas of research in data mining, but with the increasing complexity of the data, the data size is gradually increased, Common clustering algorithm has been difficult to play a role in order to cope with this situation, a number of high-dimensional data clustering, uncertain clustering, clustering algorithm based on swarm optimization have been proposed to handle these large, complex data sets.This paper introduces the research progress in the field of clustering analysis and swarm optimization, and PSO is the focus of analysis, a improved particle swarm algorithm is proposed based on the similarity, and the algorithm is used in uncertain clustering algorithm, and finally through experiment analyzed the performance of the combination algorithm. The main work of this article summarized as follows:1.The insufficiency of particle swarm algorithm is easy to premature convergence, this article propose a improved PSO algorithm by introducing a new concept of similarity between the particles to measure the degree of diversity of particle swarm, and adaptive change in the threshold adjustment means to control particle swarm algorithm convergence speed, so it can tends to the global optimum, in order to avoid falling into local optimal the iterative process in particle swarm algorithm based on similarity added Gaussian noise and other disturbances to re-adjust the position of the particles, experiment and performance analysis show that the new algorithm can effectively improve the ability of global search of the algorithm, and effectively avoid the premature convergence problem.2.The proposed optimization of PSO based on the similarity is applied to the uncertainty cluster analysis. The central idea of the new algorithm is using the improved particle swarm optimization to guide the direction of uncertain clustering, then clustering results will feedback to particle swarm optimization. In addition, with intrusion detection data experiment we analyzed performance of the combination algorithm.