Some of the well-known fuzzy clustering algorithms are based on Euclidean distance function, which can only be used to detect spherical structural clusters. Gustafson-Kessel (GK) clustering algorithm and Gath-Geva (GC) clustering algorithm were developed to detect non-spherical structural clusters. Both, of GG and GK algorithms suffer from the singularity problem of covariance matrix and the effect of initial status. In this paper, a new Fuzzy C-Means algorithm, based, on Particle Swarm Optimization and Mahalanobis distance without prior information (PSO-FCM-M) is proposed to improve those limitations of GG and GK algorithms. And we point out that the PSO-FCM algorithm is a special case of PSO-FCM-M algorithm. The experimental results of two real data sets show that the performance of our proposed PSO-FCM-M algorithm is better than those of the FCM, GG, GK algorithms.
Relation:
INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING INFORMATION AND CONTROL 5 (12B): 5033-5040 Sp. Iss. SI
