Context-Based Gustafson-Kessel Clustering with Information
Myung-Won Lee, Keun-Chang Kwak · 2014
In this paper, we propose a Context-based Gustafson-Kessel (CGK) clustering that builds Information Granulation (IG) in the form of fuzzy set. The fundamental idea of this clustering is based on Conditional Fuzzy C-Means (CFCM) clustering introduced by Pedrycz. The proposed clustering develops clusters preserving homogeneity of the clustered patterns associated with the input and output space. Furthermore, this performs the local adaptation of the distance metric to the shape of the cluster based on fuzzy covariance matrix and linguistic contexts. The experimental results reveal that the proposed clustering algorithm yields better performance in comparison with Fuzzy C-Means (FCM), GK, and CFCM clustering introduced in the previous literature for synthesis data set.