Semi-supervised clustering algorithms for web documents

Abstract

Data mining has been a significant tool in extracting hidden and useful information from large databases in various scientific and practical applications. One of the techniques is semi-supervised clustering. Semi-supervised algorithms often demonstrate surprisingly impressive performance improvements over traditional one-sided row clustering techniques by attempting to simultaneously partition both the rows and columns. In many application algorithms, partial supervision in the form of a few rows labeling information as well columns may be available to potentially increase the performance of semi-supervised clustering. In Sindhwani‟s paper, they proposed two novel semi-supervised clustering algorithms motivated respectively by spectral bipartite graph partitioning and matrix approximation formulations for co-clustering.