An improved hybrid global structure model for detection of spam webpages

Search engines retrieve information from the World Wide Web, but effectiveness is diminished by web spam techniques such as cloaking, link farming, and keyword stuffing that manipulate ranking algorithms. These activities reduce search quality and mislead users. A major challenge in combating web spam is identifying influential webpages in web graphs. Traditional centrality measures evaluate node significance but often perform poorly in weighted, directed web graphs. Although hybrid approaches, including gravity-based centrality and weighted k-shell decomposition, combine local and global structural features, they still fail to capture web graph structure. This article proposes the Improved Hybrid Global Structure Model ([Formula: see text]), which combines global structural context, weighted k-shell decomposition, and Laplacian centrality to identify significant structural characteristics of webpages and to overcome these constraints. The VIKOR multi-criteria decision-making ([Formula: see text]) framework is used to assess the efficacy of various interventions, with webpages performing as evaluation criteria and the methods under consideration serving as alternatives. Additionally, the classification capacity of the collected structural features for differentiating between spam and trustworthy webpages is validated using Support Vector Machine ([Formula: see text]). The proposed [Formula: see text] outperforms traditional centrality measures and offers an efficient framework for spam webpage detection, according to experimental evaluation on the WEBSPAM-UK2006 and WEBSPAM-UK2007 datasets.