改进朴素贝叶斯模型的复杂网络关系预测

发布时间：2018-05-07 03:24

本文选题：复杂网络 + 贝叶斯模型　；参考：《计算机工程与科学》2017年10期

【摘要】：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。
[Abstract]:Complex networks include biological information networks, scientists' cooperative networks, social networks, and so on. And excavates the potential good friend relation and so on. At present, most of the relational prediction algorithms are implemented by the similarity model of complex networks, but this type of algorithm is based on explicit network topology features and ignores the implicit information that affects the generation of relationships. In order to solve this problem, based on the naive Bayesian link prediction model (LNB), an enhanced naive Bayesian link prediction model (ELNBN) is proposed. The model models the local subgraph features of the coneighbor node by defining the probability of the co-neighbor node relationship. It effectively alleviates the independence hypothesis in LNB and realizes the quantization calculation of the contribution of the co-neighbor node relationship. Experiments on artificial data sets and real complex network data sets show that the proposed model is superior to the benchmark algorithm and other recently proposed models. At the same time, the idea of ELNB is effectively extended to other similarity algorithms based on co-neighbor nodes, which provides a new scheme for the study of this kind of model.
【作者单位】：广东工贸职业技术学院计算机工程系;华南理工大学计算机科学与工程学院;
【基金】：广东教育研究专项项目(GDJY-2014-B-B200) 广东高等职业技术教育研究会项目(GDGZ14Y037)
【分类号】：O157.5

【相似文献】