Improvements over Fuzzy clustering methods for large Datasets

Fuzzy C-means and kernel FCM-F are multi scan methods and require NC distance computations where N is the size of the dataset D, C is the number of cluster centers in the data and Kernel FCM-K also is a multi scan method requiring N2C distance computations in each iteration. For large values of N, the overall computation cost will go on increasing for these methods. The Book proposed two-step prototype based hybrid techniques to speed-up FCM, KFCM-F and KFCM-K. The proposed algorithms are called Prototype based FCM (PFCM), Prototype based KFCM- F (PKFCM-F) and Prototype based KFCM-K(PKFCM-K). Initially, few prototypes are generated from the given dataset and later the conventional methods are applied on these selected prototypes. The present work focuses on reducing the time complexities of these methods without effecting the Clustering Accuracy. The reduction in running time will make these methods work efficiently on very large data sets.