Research On K-medoids Clustering Algorithm Based On Granular Computing And Simulated Annealing

The rapid development of science and technology has brought explosive growth of data, data mining can be extracted for human life and production of useful information in confused, vast amounts of data. Cluster analysis because of its unique advantages, at this stage of data mining has become an important research direction. At the same time due to the clustering algorithm itself has some shortcomings, such as the initialization of the sensitive center searches for updates weak and the objective function easily fall into local minima, etc., need to be further refined and improved the algorithm, making clustering algorithm is more important practical value.In this paper, based on the wording of the basic theory of traditional clustering algorithms are discussed, including K-medoids clustering algorithm shortcomings as well as the latest research results, and put forward the corresponding improvement optimization. This paper follows the specific research:This article refers to the granular computing theory to solve the traditional K-medoids clustering algorithm initial cluster center initialization sensitive issue and propose an improved granular computing to traditional K-medoids algorithm based initialization process. The algorithm is tested in experiments focus on Iris data, experimental results show that: the algorithm accuracy is higher, fewer iterations.For K-means algorithm is easy to fall into local optimum drawback, based on the global use of improved simulated annealing clustering algorithm to optimize performance, increase K-means algorithm global search ability, avoid falling into local optimum. Through the sample set each dimension attribute normalized experimental results show that the algorithm can effectively suppress local convergence, and accuracy and stability are greatly improved.