Dynamic Clustering Algorithm Based on Granular Lattice Matrix Space Model

Abstract

Traditional clustering algorithm usually adopt uniform granularity. It easily leads to too fine or too coarse in clustering process. The former may divides objects into different classes which should be in one. The latter group objects into one class which should be in different. Due to it, we introduce dynamic granularity into traditional clustering algorithm. Firstly, based on research, we present granular lattice matrix space model. Then we describe problem of clustering by the new model. Finally we provide new clustering algorithm based on the new model. To testify the new algorithm, we present tests to prove its efficiency.