基于链路预测的有向互动影响力和用户信任的推荐算法

针对传统推荐算法存在忽视社交网络结构紧密强度对用户信任传递的影响和缺乏社交心理解释等问题，提出基于链路预测的有向性互动影响力和用户信任的推荐算法。首先利用融合用户偏好行为和社交圈的综合相似度识别出目标用户的相似朋友圈；其次通过结合节点引力指数和有向性影响因子获得目标用户之间的有向性互动影响力，再利用由有向性互动影响力和用户评分信任而得的综合用户信任值在目标用户的相似朋友圈中寻找出值得信任的相似用户集合，有效提高了推荐的精确性，最后产生推荐。结果表明，所提的推荐方法较之前的社会网络推荐算法在性能上具有显著提高。

Recommendation algorithm based on link prediction for directed interaction influence and user trust

Aiming at the problems that the traditional recommendation algorithm ignored the influence of the tight structure of social network structure on user trust transmission and the lack of social psychological explanation, this paper proposed a re-commendation algorithm based on link prediction for directed interaction and user trust. Firstly, it identified the similar user circle of the target user by the integrated similarity between the user preference behavior and the social circle. Secondly, it obtained the directional interaction influence between the target users by combining the node gravity index and the directed influence factor. At last, it integrated user trust value of the directional interaction influence and the user score trust to find a trustworthy similar user set in the similar circle of friends of the target user, which effectively improved the accuracy of the recommendation and finally generated the recommendation. The results show that the proposed method has a significant improvement in performance compared to the previous social network recommendation algorithm.