社会网络中的链接稳定性预测问题研究

社会网络是一个由对象和链接所构成的复杂关系型数据集．给定一个社会网络的快照，能否预测在下一个较短的时间段内其中哪些链接将会继续存在而哪些链接将会消失，这是社会网络中链接挖掘的一项新的任务．这一任务可以形式化为链接稳定性预测（Link Stability Prediction）．提出了通过打分的方法来度量链接的稳定性，并讨论了几种基于邻近度度量的链接稳定性预测方法．在一个大型的电话通信网络上进行了实验，并设计了一种用来评价链接稳定性预测准确性的机制．实验结果表明，链接的稳定性信息在一定程度上是可以通过基于邻近度度量的方法从社会网络的拓扑结构中提取出来的，其中有两三种较精炼的预测方法展现出了良好的性能．但是，这些静态的链接稳定性预测方法也有其局限性，更多动态的预测方法亟待研究．

Research on Link Stability Prediction in Social Networks

A social network is a complex relational structure constructed by entities and links. For a given snapshot of a social network, we can predict which links will keep their existence and others will disappear in the near future？ This is a new and interesting task for link mining in social networks. This task can formalized as the link stability prediction. The measurement of the stability of links, and develop approaches for link stability prediction based on proximity measures were discussed in this paper, the experiment on a large scale telephone communication network was carry out, and a mechanism to evaluate the accuracy of different link stability prediction methods was designed. The experiment suggests that information about the stability of links in a social network can be extracted from the topology of the network to some degree by using methods based on proximity measures, and two or three subtle methods perform especially well. However, there are some limitations on these static link stability prediction methods, and more dynamic methods need to be developed.