Link Recommendation For Promoting Information Diffusion In Social Networks

Previous research on link recommendation most payed attention to strengthen thesocial intercourse function, ignoring the improvement of information diffusionfunction. Link recommendation algorithms should not only focus on accuratelyestimating the proximity to user’s preferences and social relations, but also considerthat the new connections added to network by recommendation algorithms shouldmaximize information diffusion. Previous proposed a concept of node diffusiondegree on the introduction of the community detection, and combined the result ofnode diffusion degree and the result of traditional link prediction results to solve theproblem above.Research work is consists of the following three parts:1. This paper first proposes the research framework of link recommendation forpromoting information diffusion, and selects the algorithms and models for the core ofthe research framework. Then comparative analyses the existing three methods ofcomputing node diffusion degree.2. Analyses the drawback of the original node diffusion degree and proposes animproved version of node diffusion degree on disjoint community with a newcommunity pair structure.3. Analyses the information diffusion structure of overlapping community andproposes a node diffusion degree on overlapping community with node’s communitycentrality.This paper experiments on real undirected social network dataset Email-Enron,Amazon and directed social network dataset Email-EuAll, the results show that ourmethod noticeably outperforms the traditional methods in compact network in termsof promoting information diffusion.This paper pays more attention to the influence of node’s community property ofinformation diffusion. Improve and perfect the solution system of node diffusiondegree. Reduce the computational complexity greatly to apply in large-scale socialnetworks. This paper also raises two issues of concern through the experiment, whichwill help future generations to study the solution system of node diffusion degree inthe further.