Viewpoint-Driven Subspace Fuzzy C-Means Algorithm

Yiming Tang, Rui Chen, Bowen Xia

Abstract: Most of the current fuzzy clustering algorithms are sensitive to cluster initialization and do not cope well with high dimensionality. To alleviate these problems, we come up with a viewpoint-driven subspace fuzzy c-means (VSFCM) algorithm. First of all, based on the DPC (clustering by fast search and find of density peaks) algorithm, a new cut-off distance is proposed, and the cut-off distance-induced cluster initialization (CDCI) method is established as a new strategy for initialization of cluster centers and viewpoint selection. Moreover, by taking the viewpoint obtained by CDCI as the entry point of knowledge, a new fuzzy clustering strategy driven by knowledge and data is formed. We introduce the subspace clustering mode, fuzzy feature weight processing mechanism, and derive the separation formula between the clusters of viewpoint optimization. Based upon these points, we put forward the VSFCM algorithm. Finally, by comparing experiments and using multiple advanced clustering algorithms and experimenting with artificial and UCI data sets, it is demonstrated that the VSFCM algorithm has the best performance expressed in terms of five indexes.