基于排序学习的社会网络链接预测算法研究

Learning to Rank Based Link Prediction Algorithm in Social Networks

链接预测是大规模社会网络分析挖掘的重要研究内容之一,具有非常重要的应用前景.社会网络种类繁多,不同的网络链接类型往往需要不同的链接预测方法 .为了满足用户的个性化需求并提高链接预测的性能,该文提出了一种基于排序学习的社会网络链接预测算法.该算法以传统的链接预测方法为基础,通过排序学习方法对不同的排序结果进行学习,从而得到具有最大准确性的综合排序列表.在综合排序列表的构建中,在每个排序列表中设置一个滑动窗口,通过对滑动窗口的维护每次迭代选出一个全局最优值,从而使得最终的排序列表是最优的.实验表明,该文提出的算法与相关的链接预测算法相比较具有更高的预测性能,能找出一个预测最准确的排序结果 .

Link prediction is one of the most important research issues for mining and analyzing largescale social networks,and is ubiquitous in many applications.There are kinds of social networks,and different types of link need different methods for link prediction.In order to satisfy personalized user requests and improve the performance of link prediction,this paper proposed a learning to rank based link prediction algorithm in social networks.Based on traditional link prediction methods,the proposed algorithm tried to learn a final list with maximized accuracy from existing ranking list.On the constructing of the final list,we set and maintained a slide window for each ranking list,and at each iteration,chose a global optimum from all slide windows,which made sure that the final list was optimum.The experiments show that,the proposed algorithm has better performance in link prediction than related works.