Batch FCM with volume prototypes for clustering high-dimensional datasets with large number of clusters

Abstract

In this paper we present Batch Fuzzy c-Mean with Volume Prototypes algorithm suitable to cluster large high-dimensional datasets with large chosen number of existing clusters. This algorithm is much faster than the original FCM. An important part of proposed algorithm is an initialization process of the prototypes vectors, which provides better basis for finding the centers of the clusters. An another feature of the algorithm is its ability to estimate an amount of noise in the dataset. We also describe the possible application of the algorithm for the robot navigation.