Multilayer perceptrons neural network based Web spam detection application

Abstract

Web spam detection is a crucial task due to its devastation towards Web search engines and global cost of billion dollars annually. For these reasons, a multilayered perceptrons (MLP) neural network is presented in this paper to improve the Web spam detection accuracy. MLP neural network is used for Web spam classification due to its flexible structure and non-linearity transformation to accommodate latest Web spam patterns. An intensive investigation is carried out to obtain an optimal number of hidden neurons. Both Web spam link-based and content-based features are fed into MLP network for classification. Two benchmarking datasets - WEBSPAM-UK2006 and WEBSPAM-UK2007 are used to evaluate the performance of the proposed classifier. The overall performance is compared with the state of the art support vector machine (SVM) which is widely used to combat Web spam. The experiments have shown that MLP network outperforms SVM up to 14.02% on former dataset and up to 3.53% on later dataset.