An improved entropy centrality for detection of spam webpages

Centrality is a popular network analysis concept. Spam nodes in social networks are detected using centrality measures. It classifies a node as influential or anomalous. Many methods have been developed for measuring network centrality for unweighted undirected graph. This article proposed a centrality measure based on entropy centrality named Improved Entropy Centrality (IEC) for weighted directed graphs to detect spam webpages in which the graph is decomposed into subgraphs and the influence of the neighbor webpage (nodes) is analysed. The total strength of a webpage is calculated in two ways considering its direct strength(1-hop neighbors)and indirect strength(2-hop neighbors). The direct strength can be divided into two parts measured by degree entropy and weight entropy on its 1-hop neighbor webpages and indirect strength is obtained by the impact of 2-hop and 3-hop neighbor webpages. Validation has been done on different web graphs to evaluate the performance of improved entropy centrality. Additionally, the metric for distinctness is used to analyse ranking effectiveness and the entropy-weighting method is used to measure webpage weight. The experiments are carried out using the WEBSPAM-UK2006 dataset, revealing that the proposed centrality measures outperform traditional ones in distinguishing between trustworthy and spam webpages.