An Entropy based Method for Overlapping Subspace Clustering

We present subspace clustering method based on entropy and modification of Gustafson-Kessel clustering. The proposed method associates with each cluster its center, variance co-variance matrix and a weight matrix. It represents clusters and their features through gradation in membership, and hence reflects a realistic representation of clusters. Evaluation of experiments on data from UCI ae well as text with the comparative methods shows better results of proposed method.