关键词: 计算机应用, 聚类, 网格密度, 平方误差密度

Abstract:

This article proposes a mixed clustering algorithm based on grid density,which took Square Inaccuracy Density function as density evaluation criterion and avoided instability factors of classic density algorithm caused by improper setting of Eps and MinPts.It posed a kind of dynamic neighborhood radius strategy and solved the clustering deviation problem caused by classic density algorithm adopting global static neighborhood radius.In addition,it set the grid density heuristic information on all nodes in the spatial domain.It just needed computing the values of nodes in the grid domain,when constructing data structure and computing neighborhood radius,which cut the computing cost.Moreover,it just selected the representative points meeting a criterion to expand,which reduced the times of selecting and saved the running time.Finally,this article tested the algorithm in the Pendigits datasets and SEQUOIA 2000 database and the experimental results suggested that Grid Heuristic Information density clustering algorithm had a better performance than classic clustering algorithm in the clustering precision,clustering time and clustering stability.

Key words: computer application, clustering, grid density, square inaccuracy density