中文摘要：

考虑用户间影响的差异,从用户关注的兴趣点出发,及其他用户的消息或行为通过最短路径影响该用户的最大可能性,提出了基于非对称相似性的半局部拓扑指标,并将其应用于在线社交网络好友推荐。通过Facebook数据集验证了该方法,实验结果证明,考虑了非对称相似性的好友推荐算法在准确率与召回率上都明显优于其他方法,从而证实了该方法的有效性。

英文摘要：

Considering the impact of different users and the interest of the user focused,a maximum probability that the other users＇ message or behavior gave to by the shortest path was taken into account.Then,it was proposed a novel quasi-local topological similarity,and applied it to the online social network recommendation.The method outperformed in prediction and recall than other methods.The new method was verified the validity.