Original scientific paper

PNMBG: Point Neighborhood Merging with Border Grids

Renxia Wan ; College of Information Science and Technology
Jingchao Chen ; College of Information Science and Technology
Lixin Wang ; College of Information Science and Technology
Xiaoke Su ; College of Information Science and Technology

Abstract

The special clustering algorithm is attractive for the task of grouping arbitrary shaped database into several proper classes. Up to now, a wide variety of clustering algorithms designed for this task have been proposed, the majority of these algorithms is density-based. But the effectivity and efficiency still is the great challenges for these algorithms as far as the clustering quality of such task is concerned. In this paper, we propose an arbitrary shaped clustering method with border grids (PNMBG), PNMBG is a crisp partition method. It groups objects to point neighborhoods firstly, and then iteratively merges these point neighborhoods into clusters via grids, only bordering grids are considered during the merging stage. Experiments show that PNMBG has a good efficiency especially on the database with high dimension. In general, PNMBG outperforms DBSCAN in the term of efficiency and has an almost same effectivity with the later.

Keywords

Clustering; Grid clique; Point neighborhood; Border grids; Merging

Hrčak ID:

45065

URI

https://hrcak.srce.hr/45065

Publication date:

15.12.2009.

Visits: 1.143 *