An efficient ranking-centered density-based document clustering method

Abstract

Document clustering is a popular method for discovering useful information from text data. This paper proposes an innovative hybrid document clustering method based on the novel concepts of ranking, density and shared neighborhood. We utilize ranked documents generated from a search engine to effectively build a graph of shared relevant documents. The high density regions in the graph are processed to form initial clusters. The clustering decisions are further refined using the shared neighborhood information. Empirical analysis shows that the proposed method is able to produce accurate and efficient solution as compared to relevant benchmarking methods.