Abstract: Micro-blogging systems are developing tremendously and attracting a large amount of users in recent years. The tweets micro-blogging users posted provide an effective way of understanding users’ interests. Many previous researches are proposed to mine users’ interest from their posted tweets for on online friend recommendation. However, to find the information that the user really interested in is still very challenging. Furthermore, these previous works ignore the potential strong relationship between the interests and locations of users. In fact, what a user really interested in is people surrounding him in most cases. Thus, this paper proposes a geographical neighbor based friend recommendation approach, in which surrounding users who have similar interests can be detected and interact with one another in the same physical domain. By geographical neighbor based recommending, it provides users an effective way of contacting with people around them and who they may be interested in. Simulation results demonstrate the advantage of the approach compared with the traditional recommendation methods.

Key words: micro-blogging systems, friend recommendation, location, geographical neighbor

摘要： 近年来，微博的蓬勃发展吸引了大量网络用户，用户所发海量微博呈现的大数据环境成为理解用户行为的重要资源。目前，大量在线朋友推荐研究通过对微博内容分析推断用户的兴趣和喜好以进行朋友推荐，但大多数已有研究忽略了用户位置和兴趣之间的潜在关系。事实上，多数情况下用户真正感兴趣的还是他周围的人。为此，提出了基于地理近邻关系的朋友推荐方法，通过把所处位置周围兴趣爱好相似的微博用户彼此推荐，为用户提供了与周围可能感兴趣的人联系的独特渠道。仿真分析证明，与传统朋友推荐方法相比，基于地理近邻的朋友推荐具有较高的推荐性能。

关键词: 微博系统, 朋友推荐, 地理位置, 地理近邻