Data Mining Approaches for Web Spam Detection

Web spam is a serious problem for search engines because the quality of their results can be severely degraded by the presence of this kind of page. In this paper, we present an efficient spam detection system based on a classifier that combines new link-based features with language-model (LM)-based ones. We have specifically applied the Kullback-Leibler divergence on different combinations of these sources of information in order to characterize the relationship between two linked pages. In this paper, we present an efficient spam detection system based on a Hybrid clustering that combines K-means and SVM and then classified by using C4.5 with Qualified link-based features and Language Model(LM) based once. The result is an accurate system for detecting Web spam using fewer features.