Using two-stage approach to clustering

A grid-based approach to clustering is presented. Each grid is a hypercube in data space, and Appriori algorithm is used to find the representing subsets of each cluster. The two-stage procedure -first finding all representing subsets then clustering in the second stage - is found to perform well when compared with direct clustering of data. The use of the representing subsets can efficiently find the data structure of the give dataset. Consequently, the new approach can effectively overcome the parameter-sensitive problem that is encountered in most of the conventional grid-based approaches to clustering. At the same time, if a proper threshold in the new approach is chosen, the computation time to cluster a large dataset will further decrease greatly. Two experiments are used to illustrate the performances of the new proposed approach and verify its merits.