Share Email Print
cover

Proceedings Paper

An automatic clustering algorithm based on DB criterion
Author(s): Feng Zhang; Wei-ping Yang; Zhi-long Zhang; Shu-xin Zhang; Xin Ye; Jun-jun Zhang
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Inspired by a recent algorithm on clustering, we proposed an improved algorithm which combines the Davies Bouldin criterion to obtain the right number of the cluster centers automatically and output the right clusters. Davies-Bouldin criterion can describe the intra-class scatter and inter-class deviation value of the clustering result. In our algorithm, we first calculate the density and the distance of the sample points, which contain the information of the density distribution leading to the right clusters; Then, we choose two thresholds of the density and the distance to obtain the maximum number of the cluster centers; Finally, our algorithm automatically searches the right number of cluster centers through calculating the Davies-Bouldin value of every clustering result and choose the one which has the minimum Davies-Bouldin value. Experiments show that our algorithm can not only output the right clustering result when the sample points are disturbed and with special density distribution, but can also obtain the right number of the cluster centers automatically.

Paper Details

Date Published: 25 October 2016
PDF: 8 pages
Proc. SPIE 10157, Infrared Technology and Applications, and Robot Sensing and Advanced Control, 101570P (25 October 2016); doi: 10.1117/12.2244641
Show Author Affiliations
Feng Zhang, National Univ. of Defense Technology (China)
Wei-ping Yang, National Univ. of Defense Technology (China)
Zhi-long Zhang, National Univ. of Defense Technology (China)
Shu-xin Zhang, National Univ. of Defense Technology (China)
Xin Ye, National Univ. of Defense Technology (China)
Jun-jun Zhang, National Univ. of Defense Technology (China)


Published in SPIE Proceedings Vol. 10157:
Infrared Technology and Applications, and Robot Sensing and Advanced Control

© SPIE. Terms of Use
Back to Top