Analysis And Research On Rough Clustering Algorithm Based On Particle Swarm Optimization

With the rapid development of computer science technology and the change of human need, resulting in a large amount of data, data mining has become one of the hot research fields. In the massive of data field, there is much important valuable information, data mining is the technology that can tap into the human or extract useful knowledge that can effectively promote economic and social development of the ocean of the data. Clustering analysis is an important branch of data mining and it has been widely used in the production of human life.This paper mainly discusses improving clustering algorithms based on particle swarm optimization, and research aspects are as followed:Firstly, this paper proposed an improved clustering algorithm based on particle swarm optimization that is aimed to resolve the k-means algorithm shortcoming of sensitizing to the initial clustering center and easiness to fall into local optimum. The improved algorithm combined density-based and maximum minimum distance method to determine initial clustering center, so it can automatically determine the initial cluster centers,solving problem of the k-means is sensitive to the initial clustering center. The advantages of PSO’s strong global optimization ability are used to avoid K-means falling into local optimum, enhancing global search capability of k-means algorithm. By normalizing each attribute of the sample set, decreasing inertia weight by concave function, calculating the dissimilarity matrix, introducing particle swarm’s fitness variance, to optimize the hybrid algorithm further. The experimental results show that this algorithm has higher accuracy and stronger convergence capability.Secondly, the k-medoids algorithm has the disadvantage of global search ability and large amount of the iterative calculation, this paper proposed a improved rough k-medoids algorithm based on particle swarm optimization(PSO).By introducing PSO strengthen its global search ability and calculating the dissimilarity matrix of sample set to simplify coding particle swarm, the rough set theory provides a processing method of dealing with the indeterminacy problem of boundary objects. Using memorization technique to improve k-medoids iterative calculation,to reduce the complexity of the algorithm. Through testing the Iris、Mushroom data set of UCI, the new algorithm’s accuracy has improved and the time has been shortened.