Fighting WebSpam: Detecting Spam on the Graph Via Content and Link Features

摘要

We address a novel semi-supervised learning strategy for Web Spam issue. The proposed approach explores graph construction which is the key of representing data semantical relationship, and emphasizes on label propagation from multi views under consistency criterion. Furthermore, we infer labels for the rest of the unlabeled nodes in fusing spectral space. Experiments on the Webspam Challenging dataset validate the efficiency and effectiveness of the proposed method.