中文摘要：

针对基于社区划分的潜在好友推荐算法FRCD运行速度慢的问题,提出了一种基于社区划分的多线程潜在好友推荐算法MTFRCD。该算法在网络拓扑图上利用多线程技术寻找核心关系子网,以核心关系子网作为标签种子节点,使用多线程并发传播标签来发现网络拓扑图上的社区结构,利用社区发现结果在社区内部推荐潜在好友。人工网络的实验结果表明,MTFRCD相比于传统的FRCD,在性能近似的前提下具有明显的速度增长。因此,将该算法应用于真实社交网络（学者网）平台的潜在好友挖掘和推荐,根据推荐结果的评测,验证了算法具有良好的推荐效果。

英文摘要：

This paper proposed multi-thread latent-friendship recommendation algorithm based on community detection （ MT- FRCD） due to the low running speed of a method for latent-friendship recommendation based on community detection （FRCD）. Firstly, MTFRCD figured out kernel sub-networks on network topological diagram by multi-thread technology. Secondly, it regarded kernel sub-networks as seed nodes, the algorithm detected every community structure on network topological diagram by multi-thread paralleling label propagation. At the end, it recommended latent-friendships in all communities. The experiment on artificial network shows MTFRCD maintains performance and elevates running speed notably comparing to tradi- tional method FRCD. Therefore, MTFRCD can apply to the real social network（ Scholat. com） to complete latent-friendship recommendation. According to recommendation assessment, MTFRCD is able to achieve a good recommendation.