【摘要】 [目的/意义]旨在解决因微博用户弱关系结构导致难以发现潜在朋友的问题。[方法/过程]基于链路预测原理,提出预测微博双向"关注"关系的CN、RA、AA三种相似性指标加权模型,以及双向"关注"关系融合单向"关注"关系的预测模型,并在新浪微博用户关系网络数据集及推特(Twitter)用户关系网络标准数据集上进行测试。[结果/结论]提出的3种加权模型对于相似性指标推荐精度均有提升,其中RA的扩展形式最有效。更多还原

【Abstract】 [Purpose/significance]The paper is to solve the issue of being difficult to find potential friends due to weak relationship structure of weibo users.[Method/process]The paper bases on link prediction, puts forwards three kinds of weighted models using similarity indexes of CN, RA and AA for predicting bidirectional attention relationship, presents link prediction models by combining unidirectional relationship and bidirectional relationship, and takes Sina weibo and Twitter users databases to verify re sults of prediction models. [Result/conclusion]Three weighted models increase recommendation accuracy of similarity index and RA has the best result among them.更多还原