基于社交影响力的推荐算法

A Recommendation Algorithm Based on Social Influence

社交网站的兴起促使社交推荐成为了推荐系统领域的热门研究方向。微博是一类具有代表性的社交网站,其用户之间以一对多的不对等关系为主,如何为微博用户推荐潜在的关注对象是社交推荐中的一个重要研究点。文中针对微博类社交网站中用户间关系不对等的特点,结合用户间的交互行为信息,提出了一种社交影响力的计算方法,并在此基础上提出基于社交影响力的推荐算法(SIB)。该算法通过计算用户社交影响力矩阵,然后使用K个最近邻(K-Nearest Neighbor,KNN)算法找出目标用户的邻居集合,借助邻居集合帮助推荐。该算法综合考虑了微博社交网站中的两种社交关系,能有效地对微博类社交网站进行个性化推荐。通过在真实数据集上进行实验,证明该算法在微博类社交网站中的推荐效果比单纯的基于用户协同过滤(User-based Collaborative Filtering,UCF)算法有一定程度的提升。

The rise of social networking sites promotes social recommendation becoming the research hotspot in the field of recommender systems. As a representative social network,Weibo has unequal one-to-many relationship between users. How to recommend potential Weibo users concerned is an important research direction in the social recommendation. Aiming at the characteristics of unequal relationship between users in social network of Weibo, combined with the mutual behavior information between users, a computing method of social influence is proposed and on the basis,a SIB algorithm is also presented. By calculating the social influence matrix of user,this algorithm uses the KNN algorithm to find the target user set of neighbors, and helps to recommend with the aid of neighbors set. The algorithra considers the two kinds of social relations for social network in Weibo, which can effectively conduct personalized recommendation for social networking sites in Weibo. The experiment shows that the SIB algorithm can effectively improve the accuracy of recommenda- tion system in social networks compared with UCF algorithm.