Research On Improvement Of Three-Way Decision Clustering Algorithms And Applications

Owing to the overlapping clustering problems in e-commerce recommendation system,social network analysis,biomedicine and other fields.massive amounts of new data are generated every day,data relationship becomes complicated and the problem of overlap between classes becomes more prominent.Therefore,it is critical to study an algorithm that solves both incremental and overlapping problems.The emergence of three decision-making ideas provides a new solution for overlapping clustering,in which three-increment incremental clustering algorithm based on tree structure(TIOC-TWD algorithm)simultaneously solves the problem of increment and overlap.However,the TIOC-TWD algorithm does not take into account the difference in attribute importance when calculate the similarity,and the algorithm is only applicable to pure numeric attribute data and cannot handle clustering of mixed attribute datasets.To solve the above problems,this dissertation study the overlapping of incremental data and the overlap of static mixed attributes respectively.The main work of this dissertation is as follows:1.To solve overlapping clustering of incremental data,this dissertation proposed a weighted TIOC-TWD clustering algorithm which is called W-TIOC-TWD algorithm.Taking attribute importance into the distance measure.The improved distance formula of TIOC-TWD algorithm can cover the shortage of equal contribution of all attributes in the process of clustering.Moreover,The experimental results on artificial datasets and UCI datasets verify higher accuracy and stability of W-TIOC-TWD algorithm than the one of original algorithm.2.Against the overlapping clustering of static dataset with mixed attributes,this dissertation proposed MAW-SOC-TWD clustering algorithm based on theory of three-way decision.This algorithm design a new assignment method of weighted coefficient and calculating formula of similarity of SOC-TWD which is clustering algorithm of TIOC-TWD aimed at cluster static dataset.The experimental results of UCI dataset validate effectiveness of the proposed algorithm and indicate that MAW-SOC-TWD algorithm can achieve higher clustering accuracy than the comparative algorithms...