Regularized clustering for documents

Abstract

In recent years, document clustering has been receiving more and more attentions as an important and fundamental technique for unsupervised document organization, automatic topic extraction, and fast information retrieval or filtering. In this paper, we propose a novel method for clustering documents using regularization. Unlike traditional globally regularized clustering methods, our method first construct a local regularized linear label predictor for each document vector, and then combine all those local regularizers with a global smoothness regularizer. So we call our algorithm Clustering with Local and Global Regularization (CLGR). We will show that the cluster memberships of the documents can be achieved by eigenvalue decomposition of a sparse symmetric matrix, which can be efficiently solved by iterative methods. Finally our experimental evaluations on several datasets are presented to show the superiorities of CLGR over traditional document clustering methods.

Feature Selection and Kernel Learning for Local Learning-Based Clustering

“…In fact, by combining multiple kernels and exploiting the complementary information revealed from the different kernels, the LLC-mkl indeed improves the robustness and accuracy of the LLC. Moreover, the text data are believed to lie on a low-dimensional manifold because it is impossible for them to fill in the entire high-dimensional space [4]. We can also observe that the performance of LLC-mkl, which utilizes the manifold information, is consistently superior to that of NMF-mkl, which does not utilize such information.…”

A Survey of Text Clustering Algorithms

“…Matrix-factorization provides a natural way of achieving this goal. It has also been shown both theoretically and experimentally [33,93] that the matrix-factorization technique is equivalent to another graph-structure based document clustering technique known as spectral clustering. An analogous technique called concept factorization was proposed in [98], which can also be applied to data points with negative values in them.…”

Enhancing Text Document Clustering Using Non-negative Matrix Factorization and WordNet

“…However, this method cannot be kernelized because the NMF must be performed in the original feature space of the data points. Wang et al [9] used clustering with local and global regularization (CLGR), which uses local label predictors and global label smoothing regularizers. They achieved satisfactory results because the CLGR algorithm uses fixed neighborhood sizes.…”

“…They achieved satisfactory results because the CLGR algorithm uses fixed neighborhood sizes. However, the different neighborhood sizes cause the final clustering results to deteriorate [9].…”

“…Internal knowledge-based document clustering determines the inherent structure of a document set by using a factorization technique [4][5][6][7][8][9]. These methods have been studied intensively, and although they have many advantages, the successful construction of semantic features from the original document set remains limited with respect to the organization of very different documents or the composition of similar documents [10].…”