基于核心用户对发现的微博好友推荐算法

Micro-blog friend recommendation algorithm based on finding of core user-pairs

现有的微博好友推荐算法没有充分考虑网络关系结构,导致发现的邻居群合理性较差。针对这个问题,围绕微博用户群聚规律和社交网络特点展开研究,提出一种基于核心用户对发现的微博好友推荐算法;该算法首先将任意两个具有相互关注关系的用户封装成用户对的形式并计算各用户对之间的交互行为相似度,然后通过密度和距离两个参数发现核心用户对以及划分合理的邻居类簇,最后根据制定的推荐规则向用户进行好友推荐。结果表明,相比传统的协同过滤方法,该算法明显提高了微博好友推荐的精度,核心用户对发现、类簇的合理划分以及推荐规则的制定能够缓解数据稀疏和冷启动带来的问题。

The existing recommendation algorithms often do not fully consider network relationship structure, leading to find irra- tional neighbors. Therefore,according to micro-blog users＇ clustering rules and the characteristics of social network, an algorithm for micro-blog friend recommendation based on the finding of core user-pairs was proposed. Firstly, any two users who had friendship between each other were put into a user-pair, and the interaction similarities of all user-pairs were computed. Secondly, core user-pairs were found through two parameters, density and distance, and rationally classified into clusters. Finally, micro-blog friends were recom-mended to users according to recommendation rules. The results show that, compared with the traditional collaborative filtering methods, this method can greatly improve the precision of micro-blog friend recommendation, and the finding of core user-pairs, reasonable cluster classification and the recommendation rules can ease the problems of data sparseness and cold start.