Information Recommendation Between User Groups in Social Networks

Abstract

Information recommendation between different user groups has recently received a lot of attention in the information service community. However, we find that obtaining the exact optimal recommendation solution is an NP-hard problem. Based on the above finding, in this paper, we present an efficient method achieving approximate optimal recommendation solution (AAORS) to reduce this NP-hard problem to an equivalent extended Steiner tree problem and obtain the approximate optimal recommendation solution appIRS in polynomial time. We theoretically prove that the global trust value of appIRS is at least 63 % of that obtained for the exact optimal solution optIRS. Moreover, in real applications, based on a computed index of reputation gain, we also adjust the recommendation solution produced by the AAORS method in polynomial time and obtain the optimal recommendation solution which satisfies the global reputation constraint. The detailed theoretical analyses and extensive experiments demonstrate that our proposed methods are both efficient and effective.