A statistical information-based clustering approach in distance space

• Published in: Journal of Zhejiang University. A. Science Vol. 6; no. 1; pp. 71 - 78
• Main Author: 岳士弘 李平 郭继东 周水庚
• Format: Journal Article
• Language: English
• Published: Institute of Industrial Process Control, Zhejiang University, Hangzhou 310027, China%Department of Mathematics, Yili Teacher's College, Yining 835000, China 01.08.2005
• Subjects: DBSCAN; 数据库; 空间数据; 统计信息; 聚类分析; DBSCAN algorithm; Statistical information; Threshold
• ISSN: 1673-565X; 1862-1775
• DOI: 10.1631/BF02842480

Clustering, as a powerful data mining technique for discovering interesting data distributions and patterns in the underlying database, is used in many fields, such as statistical data analysis, pattern recognition, image processing, and other business applications. Density-based Spatial Clustering of Applications with Noise (DBSCAN) (Ester et al., 1996) is a good performance clustering method for dealing with spatial data although it leaves many problems to be solved. For example,DBSCAN requires a necessary user-specified threshold while its computation is extremely time-consuming by current method such as OPTICS, etc. (Ankerst et al., 1999), and the performance of DBSCAN under different norms has yet to be examined. In this paper, we first developed a method based on statistical information of distance space in database to determine the necessary threshold. Then our examination of the DBSCAN performance under different norms showed that there was determinable relation between them. Finally, we used two artificial databases to verify the effectiveness and efficiency of the proposed methods.