User recommendation for promoting information diffusion in social networks

Online social networks mainly hold two functions: social interaction and information diffusion. Most of existing user recommendation studies only focused on enhancing the social interaction function, but ignored the problem of how to strengthen the information diffusion function. Aiming at this drawback, this paper introduces the concept of user diffusion degree, then combines it with traditional recommendation methods for reranking recommended users. Specifically, we propose two user diffusion degree calculation methods, node granularity algorithm and community granularity algorithm, which fully exploit the community attributes of users. Experimental results on Email and Amazon datasets under Independent Cascade Model illustrate that our methods noticeably outperform traditional recommendation methods in terms of promoting information diffusion. We also find that node granularity algorithm performs better in spares networks, while community granularity algorithm is more suitable for dense networks.