Classification of High-dimensional Data Clustering Based on Rules Mining Research

摘要

on the classification of high-dimensional data clustering analysis, traditional similarity index and dimension reduction based on clustering analysis method is hard to avoid "dimension disaster" problem or sampling errors. Therefore, on the basis of choosing the most sub space of the rough set theory, the article directly make a research of the classification of high dimensional data clustering theory mode through to the "equivalence relation" rule mining. Besides, through the China mobile company five cities sampling data of the loss of cell phone users, we has carried on the empirical test and a better clustering results are obtained. In the comparison of KMeans, Two-step and Kohonen methods of clustering, In this paper, classification of high-dimensional data clustering method based on equivalence relation in the type definition, rule mining, the number of iterations which has unique advantages and variable selection.