Reverse hierarchical variable clustering on database accelerated algorithm

This paper presents one reverse hierarchical variable clustering on database accelerated algorithm. The present method can both enhance the operation speed in large scale sparse database and save the storage space, simultaneously keep the inherent structure of the database. The main thought of the paper is using reverse variable cluster method to produce a projection clustering database. Many attributes are possible gathered under one projection clustering variable, so the attributes number can be reduced obviously. The present method can compress the sparse database without loss, which improving the speed and saving the storage space at the same time. The experiment indicate that along with the change of the original attribute number of each cluster variable represents, the algorithm can enhance efficiency from 1.17 to 5.04 times, and the attribute number can reduce approximately 88%.