Abstract: Concerning the problems of current social networking friends recommended methods, such as no obvious user interest and poor correlation between the users, a collaborative filtering algorithm was proposed based on common users and similar labels. The common concerned users were extracted as joint project data, and the custom labels were added to reflect the users' interest. Then the semantic similarity of the labels was calculated to expand the sparse matrix and improve the collaborative filtering recommendation. The experimental results show that, compared with the traditional collaborative filtering algorithm with single index, the proposed algorithm can better reflect the users' interest, and has significant improvement in the recommended accuracy and the average accuracy.

Key words: label, social network, collaborative filtering, users recommendation, semantic similarity

摘要： 针对目前的社交网络好友推荐方法用户兴趣不明显、用户之间相关性较差等问题，提出一种基于共同用户和相似标签的协同过滤算法。抽取共同关注用户作为共同项目，加入体现用户兴趣的自定义标签数据，并对标签进行相似度计算处理，以扩充稀疏矩阵，改善协同过滤推荐方法。实验结果表明，与单指标的协同过滤推荐算法相比，基于共同用户和相似标签的好友推荐方法更好地体现了用户兴趣，同时在推荐准确率和平均准确率上都有较大提高。

关键词: 标签, 社交网络, 协同过滤, 用户推荐, 语义相似度