文献类型：期刊文献

英文题名：Constructing friendship in social networks for precise peer influence marketing by consensus link prediction algorithm

作者：Li, Shugang[1];Cai, Nannan[1];Yu, Zhaoxu[2]

机构：[1]Shanghai Univ, Sch Management, Shanghai 200444, Peoples R China;[2]East China Univ Sci & Technol, Dept Automat, Shanghai 200237, Peoples R China

年份：2020

卷号：79

期号：11-12

起止页码：7649

外文期刊名：MULTIMEDIA TOOLS AND APPLICATIONS

收录：;EI(收录号：20200207995867);WOS:【SSCI(收录号:WOS:000505356600015)，SCI-EXPANDED(收录号:WOS:000505356600015)】；

语种：英文

外文关键词：Friendship construction; CLPA; PIM; Social networks

摘要：In peer influence marketing (PIM), friendship can significantly influence customers' purchase decisions. While so far, existing researches ignore friendship construction among dissimilar influencers and target users in social networks, which help greatly to expand the friend count of influencers. Therefore, this study proposed a model to construct friendship among dissimilar users by smartly recommend friends to them. Specifically, for the sake of overcoming the influence of the scale-free topology on prediction accuracy, consensus link prediction algorithm (CLPA) is proposed to predict possible friendship between users in social networks by adaptively constructing the composite similarity index (SI) for specific user pairs. Based on the prediction results of CLPA, hill climbing algorithm (HCA) is developed to construct friendship among dissimilar users by increasing their similarity in social relationship. To share more consistency with other SIs in calculating user pair similarity, CLPA smartly combines the multiple SIs into one composite SI by using their total mean rough classification uniformity (TMRCU) and taboo search algorithm (TSA), where TMRCU is used to measure the consistency of various SIs in classifying the similar users and TSA is adopted to optimize the weights of SIs with different TMRCU. Furthermore, in CLPA, cluster consistency of user pairs is developed to identify the similar user pairs in the light of the high clustering consensus between the composite SI and other SIs. Finally, the experimental results in real social networks show that the proposed method is promising for precise PIM.