改进朴素贝叶斯模型的复杂网络关系预测

An enhanced naive Bayesian relationship prediction model in complex networks

复杂网络包括生物性信息网络、科学家合作网络、社交关系网络等,研究复杂网络的关系预测问题有助于预测蛋白质相互关系,发现科学家合作关系,以及挖掘潜在好友关系等。目前,绝大多数关系预测算法由复杂网络的相似度模型实现,但该类型算法基于显式的网络拓扑特征构建,忽视了影响关系生成的隐含信息。针对这一问题,在朴素贝叶斯链接预测模型(LNB)基础上提出了一种加强(Enhanced)朴素贝叶斯链接预测模型(ELNB),该模型通过定义共邻节点关系概率对共邻节点构成的局部子图特征进行建模,有效缓解了LNB中的独立性假设,实现了共邻节点关系贡献的量化计算。在人工数据集和真实复杂网络数据集上的实验表明,本文提出的模型优于基准算法和其他新近提出的模型。同时,把ELNB的思想有效地拓展到其他基于共邻节点的相似度算法中,为该类模型的研究提供一种新的方案。

Complex networks include biological information networks, collaboration networks and social networks. Studying the relationship prediction of complex networks helps predict relationship be-tween proteins, find out cooperation relationship among scientists, as well as mine potential social net-works. Currently, most relationship prediction algorithms are realized by similarity-based models, h o w -ever ,this type of algorithms based on network topology feature are explicitly constructed, which ignore latent information behind generated relationship. T o solve this problem, w e propose an Bayesian relation prediction model （ELNB ） , which defines a conditional probability to model the localsub-graph structure. It can effectively alleviate the independence assumption of L N B and realize a quan-titative calculation of neighbors contribution. Experiments on artificial datasets and real that the proposed model is better than the baselines and some recently proposed models. Meanwhile, the idea of ELNB can be extended to other similarity algorithms based on c o m m o n neighbor nodes, which provides a new method for the research of such kind of model.