J4

• Original Articles • Previous Articles     Next Articles

New method to improve DBSCAN clustering algorithm quality

FENG Shao-rong1,2;XIAO Wen-jun1
  

  1. (1. School of Computer Science and Engineering, South China University of Technology, Guangzhou 510640, China;
    2. College of Information Science and Technology, Xiamen Univ., Xiamen 361005, China)
  • Received:1900-01-01 Revised:1900-01-01 Online:2008-06-20 Published:2008-05-30
  • Contact: FENG Shao-rong E-mail:shaorong@xmu.edu.cn

Abstract: There are three problems along with the Density Based Spatial Clustering of Applications with Noise(DBSCAN) Clustering Algorithm: input sensitivity, desire for too much memory space and the effect of nonuniform data. To solve these problems, a fast Data Partition DBSCAN using Genetic Algorithm(DPDGA) Algorithm is developed which considerably improves the cluster quality. First, the Genetic Algorithm is used to improve the K-means Algorithm to get the initial clustering center. Second, data is partitioned and the DBSCAN Algorithm is applied to cluster partitions. Finally, all clustered result sets are merged. Simulation experiments indicate that the DPDGA Algorithm works well to solve these problems and that both the efficiency and the cluster quality are better than those of the original DBSCAN Algorithm.

Key words: clustering algorithm, genetic, data partition, density

CLC Number: 

  • TP301.6