A New Clustering Validity Index for Evaluating Arbitrary Shape Clusters

When doing clustering analysis it always needs a clustering validity index to evaluate if the present clustering scheme can reflect the real natural structure of the dataset. The clusters founded by the clustering algorithm can be of arbitrary shape, but the exiting validity indices can only assess...

Full description

Saved in:
Bibliographic Details
Published in2007 International Conference on Machine Learning and Cybernetics Vol. 7; pp. 3969 - 3974
Main Authors Shang Liu, Ya-Lou Huang
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2007
Subjects
Algorithm design and analysis
Clustering algorithms
Clustering analysis
Costs
Cybernetics
Density-based clustering algorithm
Machine learning
MST
Partitioning algorithms
Shape
Testing
Tree graphs
Validity index
Visualization
Online AccessGet full text

Cover

More Information
Summary:When doing clustering analysis it always needs a clustering validity index to evaluate if the present clustering scheme can reflect the real natural structure of the dataset. The clusters founded by the clustering algorithm can be of arbitrary shape, but the exiting validity indices can only assess the validity of convex clusters. To solve this problem a new validity index CompSepa is proposed in this paper, which can evaluate a cluster scheme including both non-convex and convex clusters, and the validity index CompSepa is computed by the minimum-cost spanning tree (MST) of the objects of clusters. Experiments show that the new validity index can evaluate the clustering scheme correctly and effectively.
ISBN:1424409721
9781424409723
ISSN:2160-133X
DOI:10.1109/ICMLC.2007.4370840