一种改进的异构链路协同预测算法研究

发布时间：2018-06-04 04:19

本文选题：异构信息网络 + 相关性指标　；参考：《计算机工程与应用》2017年15期

【摘要】：现有的链路预测方法仅考虑单种链路类型预测或多种链路类型的独立预测,经常使得预测结果不够准确。为此,研究了异构信息网络中多种链路类型的协同预测问题。根据源节点的相似节点和目标节点的相似节点之间的当前链路信息,提出了同质连接原理,设计了一种针对不同类型节点的相关性指标,用于描述不同类型节点间的链路存在概率,并将其与传统的邻近性指标相结合拓展到异构链路预测中。然后,将异构信息网络中的被标记数据和无标记数据融合起来,提出一种异构链路协同预测算法(Heterogeneous Collective Link Prediction,HCLP),通过获得不同类型链路间的各种复杂关系,结合互补性预测信息,实现多种链路类型的协同预测。基于真实场景的实验结果表明,所提的链路协同预测方法可有效提升异构信息网络的链路预测性能。
[Abstract]:The existing link prediction methods only consider the single link type prediction or the independent prediction of multiple link types, which often make the prediction results inaccurate. Therefore, the cooperative prediction of multiple link types in heterogeneous information networks is studied. According to the current link information between the similar node of the source node and the similar node of the target node, the principle of homogenous connection is proposed, and a correlation index for different types of nodes is designed. It is used to describe the link existence probability between different types of nodes and to extend it to heterogeneous link prediction by combining it with the traditional proximity index. Then, combining tagged data and unlabeled data in heterogeneous information networks, a heterogeneous Collective Link prediction algorithm is proposed. By obtaining various complex relationships between different types of links and combining complementary prediction information, a heterogeneous Collective Link prediction algorithm is proposed. Cooperative prediction of multiple link types is realized. Experimental results based on real scene show that the proposed link cooperative prediction method can effectively improve the link prediction performance of heterogeneous information networks.
【作者单位】：长沙医学院信息工程学院;南京大学计算机科学与技术系;
【基金】：湖南省教育厅研究课题(No.16C0184) 国家自然科学基金重点项目(No.61432008)
【分类号】：O157.5;TP301.6

【参考文献】