Distribution Based Cluster Structure Selection

Traditional clustering integration algorithm only focuses on how to put the clustering results from multiple datasets into class label with traditional method.Different from the traditional clustering integration algorithm,distribution based cluster structure ensemble framework investigates the problem of how to select the suitable cluster structures in the ensemble which will be summarized to a more representative cluster structure.Unfortunately,the original cluster structure integration algorithm has some limitations:(1)a large amount of information will be lost when using the clustering center take the place of cluster structure.(2)Without considering cluster structure selection problem,which that no all cluster structure helps to unified cluster structure.To solve these limitations,this paper investigates a new Gaussian mixture distribution based cluster structure integration algorithm,which using Gaussian mixture distribution model to describe the clustering structure,using Bhattacharyya distance to measure the similarity between two cluster structure.The new method has designed few diversity and cohesion based cluster structure selection strategies,and used the cluster structure based consensus function to refine the unified cluster structure.The new method will be beneficial to improve the performance of clustering structure integration algorithm and provides new way for clustering structure integration algorithm in big data mining.This paper has compared the new method with traditional clustering algorithm and Clustering integration algorithm and to study the variable's contribution to the final result.