A Robust Algorithm for Fuzzy Document Clustering

Abstract

In many applications of document clustering, a document may include multiple topics and thus may relate to multiple categories at the same time. Most of the existing subspace clustering algorithms can only perform hard clustering on document collections. In this paper, a fuzzy algorithm named R-FPC is introduced for document clustering. The algorithm discovers soft partitions of a data set in the soft subspaces of the data space. Using the proposed R-Greedy initialization method, R-FPC can always generate stable clustering results with competitive accuracy. The experiments are conducted on some widely used corpuses and the results have shown effectiveness and robustness of the proposed methods.