摘要： 在异质信息网络中,节点与链接的异质特性和复杂性会增加链路预测的难度。针对该问题,提出一种采用综合拓扑特征表示的链路预测方法。对不同的元路径根据异质和同质信息网络拓扑特征获得异质和同质数据,将逻辑回归模型作为链路预测模型,并综合拓扑特征进一步提高预测准确率。在异质的movielens电影数据集上进行实验,结果表明,该方法可有效提高异质信息网络演员合作关系的链路预测性能。

关键词: 异质信息网络, 同质信息网络, 分类算法, 链路预测, 元路径

Abstract: In heterogeneous information network,heterogeneous characteristics and complexities of the nodes and links increase the difficulty of link prediction.Aiming at this problem,this paper proposes a link prediction method by using synthetic topology characteristics.For different meta-paths,it uses the topology characteristics of heterogeneous and homogeneous information network to obtain heterogeneous and homogeneous data.It uses Logistic Regression(LR) model as a link prediction model,and integrates topology characteristics to further improve the prediction accuracy.Experimental results in real heterogeneous movielens film data sets show that the proposed method can improve the link prediction performance of actor cooperation relationship in heterogeneous information network.

Key words: heterogeneous information network, homogeneous information network, classification algorithm, link prediction, meta-path

中图分类号: 

• TP18