通信学报 ›› 2021, Vol. 42 ›› Issue (7): 61-69.doi: 10.11959/j.issn.1000-436x.2021055

• 学术论文 • 上一篇    下一篇

基于高阶路径相似度的复杂网络链路预测方法

顾秋阳1,2, 吴宝1,2, 池仁勇1,2   

  1. 1 浙江工业大学管理学院,浙江 杭州 310023
    2 浙江工业大学中国中小企业研究院,浙江 杭州 310023
  • 修回日期:2021-01-27 出版日期:2021-07-25 发布日期:2021-07-01
  • 作者简介:顾秋阳(1995− ),男,浙江杭州人,浙江工业大学博士生,主要研究方向为智能信息处理、数据挖掘、中小企业高质量发展等
    吴宝(1979− ),男,浙江金华人,博士,浙江工业大学研究员、博士生导师,主要研究方向为复杂网络链路预测、金融信用风险控制与中小企业发展
    池仁勇(1959− ),男,浙江温州人,博士,浙江工业大学教授、博士生导师,主要研究方向为复杂网络链路预测、中小企业智能信息管理与创新创业
  • 基金资助:
    国家自然科学基金资助项目(71173194);国家社科基金资助项目(20VYJ073);国家社科基金资助项目(17ZDA088);浙江省社科规划重点基金资助项目(20NDJC10Z)

Link prediction method based on the similarity of high path

Qiuyang GU1,2, Bao WU1,2, Renyong CHI1,2   

  1. 1 School of Management, Zhejiang University of Technology, Hangzhou 310023, China
    2 China Institute for Small and Medium Enterprises, Zhejiang University of Technology, Hangzhou 310023, China
  • Revised:2021-01-27 Online:2021-07-25 Published:2021-07-01
  • Supported by:
    The National Natural Science Foundation of China(71173194);The National Social Science Foundation of China(20VYJ073);The National Social Science Foundation of China(17ZDA088);The Social Science Planning Key Foundation of Zhejiang Province(20NDJC10Z)

摘要:

针对目前链路预测方法普遍存在精度不高、效率低等问题,提出了基于高阶路径相似度的复杂网络链路预测方法。首先,利用路径作为判别特征对复杂网络中的缺失链接进行预测,以实现资源的有效分配,并通过惩罚公共近邻对信息泄露进行限制。其次,将高阶路径作为判别特征,对种子节点对间的可用长路径实施惩罚。最后,利用多个真实复杂网络数据集进行数值算例。实验结果表明,与其他基线方法相比,所提方法具有更优的精度与效率。

关键词: 高阶路径相似度, 复杂网络, 链路预测, 相似性度量, 公共近邻

Abstract:

For the problem that the existing link prediction method has many problems, including low accuracy and low efficiency, a method of high-order path similarity link prediction was proposed.Firstly, the path was used as the judging feature to predict missing links in complex networks, which could make resource allocation more effective and restricts information leakage by punishing public neighbor pairs.Secondly, by using high order paths as judging features, the available long paths between seed nodes would be punished.Finally, several real complex network datasets were used for numerical examples calculation.Experimental results show that the proposed algorithm is more accurate and efficient than other baseline methods.

Key words: similarity of high path, complex network, link prediction, similarity measurement, common neighbor

中图分类号: 

No Suggested Reading articles found!