A Fuzzy Document Clustering Model Based on Relevant Ranked Terms

Abstract

The web today is a growing universe of vast amounts of documents. Clustering techniques help to enhance information retrieval and processing huge volume of data, as it groups similar documents into one group. The relevant feature identification from a high-dimensional data is one of the challenges in text document clustering. We propose a sentence ranking approach which finds out the relevant terms in the documents so as to improve the feature identification and selection. Preserving the correlation between terms in the document, the document vectors are mapped into a lower dimensional concept space. We used k-rank approximation method which minimizes the error between the original term-document matrix and its map in the concept space. The similarity matrix is converted into a fuzzy equivalence relation by calculating the max-min transitive closure. On this, we applied fuzzy rules to efficiently cluster the documents. Our proposed method has shown good accuracy than previously known techniques.