基于子图特征的科学家合作网络链路预测

提出了多种基于子图结构特征的新特征,构建了基于节点重要性、基于节点共同邻居、基于边共同邻居、基于邻居子图和基于边子图五类特征,并将这五类特征中的多种特征分别作为特征输入,运用机器学习的方法,实现科学家合作网未来合作关系的预测。研究中发现,基于边子图特征的链路预测准确率最好。此外,研究中运用基于模型的特征排序和最大信息系数特征选择方法分析类内特征的影响力以及相互关系,通过机器学习算法的分类模型进行链路预测。该方法能够有效地揭示网络类内特征在预测中的重要性和相关性,有利于发现影响力大的特征和冗余特征。

Link Prediction of Cooperative Networks Based on Subgraph Features

This paper proposes a variety of new features based on the subgraph structure features. It constructs five types of features based on node importance, common neighbors based on edges, common neighbors based on edges, subgraphs based on neighbors, and subgraphs based on edges. Various features in the model are used as feature inputs, and the machine learning method is used to realize the prediction of the future cooperative relationship of the Scientist Cooperative Network. It is found in the research that the accuracy of link prediction based on edge subgraph features is the best. In addition, in the study, model-based feature ranking and maximum information coefficient feature selection methods are used to analyze the influence and interrelationship of intra-class features, and use the classification model of machine learning algorithms to perform link prediction. This method can effectively reveal the importance and correlation of intra-class features in prediction, and is conducive to discovering influential features and redundant features.