一种采用社团信息的链接预测方法

Link prediction method using community information

链接预测研究如何利用网络中已有的信息预测可能存在的关系链接,目前已成为数据挖掘领域的热点研究问题之一。社会网络中普遍存在社团结构,社团对链接的形成有重要的影响,但在大多数链接预测方法中未得到深入研究。针对这一现象提出一种新的链接预测方法,采用社团信息改进节点对样本的描述,并在监督学习框架中学习和预测。在现实数据集Facebook和ACF中的实验结果表明,加入社团信息的链接预测方法获得了更高的准确率。

Link prediction studies how to use the existing information in the network to predict the potential relationship between unlinked nodes, and it has been one of the hottest research problems in data mining. Community structures exist prevalently in social networks, they have significant impact on forming links. However, people have not thoroughly studied this link prediction problem. To deal with the above-mentioned phenomenon, this paper proposed a novel link prediction method. It not only improved the structural description of node pairs in network by adding community information, but also used supervised learning method to proceed link prediction. Experimental results on Facebook and ACF datasets show that it can raise the accuracy of link prediction by using community information in the network.