The Research And Application Of Subspace Clustering Algorithms

With the increasing market competition and the development of information technology,enterprise management turn to customer-centered mode from product-centered mode step by step.It is necessary for enterprises to provide customers with personalized services according to their different requirements.Therefore,enterprises have to conduct customer segmentation by the features of different customers.The conventional customer segmentation methods include practice based classifications or statistic based partitions which are unfit for current market needs.By virtue of the high-dimensional features of customer properties,this dissertation chooses sub-space clustering technology and applies the improved clustering algorithm to segment customers.The results are scientific and reasonable and supplies dependable gist for market decision.The main works of the dissertation are listed as follows:(1) In view of the great time complexity and low clustering precision in the conventional CLIQUE algorithm,the dissertation proposes OptCLIQUE algorithm which refines the grid cells locating at the group edges in grid space and improves clustering precision by grid partition.OptCLIQUE is different from other grid partition clustering algorithms by reprocessing the spare grid cells which are neighbors of dense ones and then it re-confirms the bound of grid cells.(2) OptCLIQUE algorithm uses grid threshold function to determine the type of grid cells,in this way it can determine parameters automatically instead of manually.The algorithm efficiency is improved by compressing transactions,deducing the table joining times and scanning data times.(3) Simulated experiment results show that OptCLIQUE algorithm is better than CLIQUE and DBSCAN in both clustering effects and performance.Applying OptCLIQUE algorithm to customer segmentation of telecommunication industries will make them master their customer dynamic information better.It provides strong support for them to maintain customer information and mine the potential customers.The practical application results also show that OptCLIQUE algorithm is efficient in customer segmentation.