Title :
A new efficient validity index for fuzzy clustering
Author :
Bouguessa, Mohamed ; Wang, Sheng-rui
Author_Institution :
Dept. of Comput. Sci., Sherbrooke Univ., Que., Canada
Abstract :
A new validity index for fuzzy clustering, which works well in the presence of overlapping clusters that differ in shape, density and orientation, is introduced. A quantitative evaluation of our new validity index in conjunction with seven well-known validity indices is given. For this purpose, we use the IRIS data set and generated truthed data sets with prescribed degrees of overlap between clusters. These truthed data sets provide a precise way to evaluate the ability of the validity indices in terms of sensitivity to cluster overlapping.
Keywords :
fuzzy set theory; pattern clustering; statistical analysis; IRIS data set; fuzzy clustering; quantitative evaluation; statistical analysis; truthed data sets; validity index; Clustering algorithms; Computer science; Covariance matrix; Design for experiments; Fuzzy sets; Iris; Maximum likelihood estimation; Partitioning algorithms; Shape;
Conference_Titel :
Machine Learning and Cybernetics, 2004. Proceedings of 2004 International Conference on
Print_ISBN :
0-7803-8403-2
DOI :
10.1109/ICMLC.2004.1382092
