基于最优路径的异构学术网络重要文献探测方法

摘要

基于同构学术网络引文网络最优路径研究，提出一种异构网络最优路径算法及两种路径重要性的评价指标，解决了现有同构网络最优路径算法不能应用于真实学术网络研究的问题。以微软学术迁移学习领域数据为数据集，从异构学术网络最优路径探测迁移学习领域重要文献与引文网络主路径探测迁移学习领域重要文献的相似性及迁移学习领域发展历程两方面验证了基于最优路径的异构网络重要文献探测方法的正确性。实验结果表明，该算法和指标更符合真实学术网络。%Based on the research of homogeneous academic network and citation network optimal-path, we proposed a heterogeneous network optimal-path algorithm and two evaluation indexes of the importance of path,and solved the problem that homogeneous network optimal-path algorithm could not be applied to the real academic network.Taking the transfer learning domain data from Microsoft as dataset,we verified the correctness in heterogeneous academic network optimal-path detection of important literatures and their similarity in transfer learning domain and the development of transfer learning.The experimental result shows that the algorithm and indexes are more consistent with the real academic network.