Truncated TrustRank: Propagating Trust Based on Target Node

Regarding impact of the ranking of web pages on financial goals, web search became a very important issue in web. Subsequently, spam pages appear, which try to deceive search engines. In this study, a framework has been proposed for detecting spam pages. In this framework, at first statistical analysis on the large data set of web pages is done, after that, by using new hybrid algorithm, trust score for both spam and nonspam pages are calculated. Presented framework using some features to classify pages as spam or nonspam based on link structure of web. Based on classifier results in TrustRank algorithm, every node in the first level of trusted seed receive the deserve trust. Worthiness of a node is specified by classifier results. Experiments have been done on sample of Farsi web graph with 800,000 nodes. By using proposed framework and propagating trust based on nature of target node, number of spam nodes which receive high trust scores is less than TrustRank algorithm.