Search

Link Spam has indentified as one of the major obstacles for link-based ranking algorithms of modern search engine since it intently constructs hyperlink structure to help some poor-content pages obtaining undeserved high rank. This problem is even worse with the advent of wikis, blogs and forum that are rich in links. Existing works on link spam are mainly focused on link spam detection by extracting some special link structures (e.g. clique, tight bipartite etc.). However, link spam structures could have many variations and easily make the existing detection methods ineffective. In this paper, we tackle the problem of link spam from a more fundamental viewpoint–˵noisy link″ analysis. First of all, how ˵non-voting″ hyperlinks affect the quality of ranking is investigated, and then based on this investigation, an approach to detect and process ˵noisy link″ both effectively and automatically is proposed. We also compare our work with two other related works (TrustRank and Site-level Noise removal) on two real web datasets. The experimental results demonstrate that the proposed ˵noisy link″ analysis is very effective on both spam page filtering and final ranking improvement. [ABSTRACT FROM AUTHOR]

Social network often serves as a medium for the diffusion of ideas or innovations. The problem of influence maximization which was posed by Domingos and Richardson is stated as: if we can try to convince a subset of individuals to adopt a new product and the goal is to trigger a large cascade of further adoptions, which set of individuals should we target in order to achieve a maximized influence? In this work, we proposed a potential-based node selection strategy to solve this problem. Our work is based on the observation that local most-influential node-selection adopted in many works, which is very costly, does not always lead to better result. In particular, we investigate on how to set two parameters(θv and buv) appropriately. We conduct thorough experiments to evaluate effectiveness and efficiency of the proposed algorithm. Experimental results demonstrate that our approximation algorithm significantly outperforms local-optimal greedy strategy. [ABSTRACT FROM AUTHOR]