Clonal selection based fuzzy C-means algorithm for clustering

In recent years, fuzzy based clustering approaches have shown to outperform state-of-the-art hard clustering algorithms in terms of accuracy. The difference between hard clustering and fuzzy clustering is that in hard clustering each data point of the data set belongs to exactly one cluster, and in fuzzy clustering each data point belongs to several clusters that are associated with a certain membership degree. Fuzzy c-means clustering is a well-known and effective algorithm, however, the random initialization of the centroids directs the iterative process to converge to local optimal solutions easily. In order to address this issue a clonal selection based fuzzy c-means algorithm (CSFCM) is introduced. CSFCM is compared with the basic Fuzzy C-Means (FCM) algorithm, a genetic algorithm based FCM (GAFCM) algorithm, and a particle swarm optimization based FCM (PSOFCM) algorithm.