《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (4): 1131-1138.DOI: 10.11772/j.issn.1001-9081.2025040448

• 数据科学与技术 • 上一篇    下一篇

基于高阶特征聚合的时间序列异常检测方法

索逸凡1, 刘松华2(), 郝秋智2   

  1. 1.太原理工大学 计算机科学与技术学院(大数据学院),山西 晋中 030600
    2.太原理工大学 软件学院,山西 晋中 030600
  • 收稿日期:2025-04-25 修回日期:2025-06-19 接受日期:2025-06-25 发布日期:2025-06-30 出版日期:2026-04-10
  • 通讯作者: 刘松华
  • 作者简介:索逸凡(1997—),男,山西吕梁人,硕士研究生,CCF会员,主要研究方向:时间序列异常检测、数据挖掘
    郝秋智(2000—),男,山西吕梁人,硕士研究生,主要研究方向:数据挖掘、异常检测。
  • 基金资助:
    国家自然科学基金资助项目(62176177);山西省自然科学基金资助项目(202203021211144)

Time series anomaly detection method based on high-order feature aggregation

Yifan SUO1, Songhua LIU2(), Qiuzhi HAO2   

  1. 1.College of Computer Science and Technology (College of Data Science),Taiyuan University of Technology,Jinzhong Shanxi 030600,China
    2.School of Software,Taiyuan University of Technology,Jinzhong Shanxi 030600,China
  • Received:2025-04-25 Revised:2025-06-19 Accepted:2025-06-25 Online:2025-06-30 Published:2026-04-10
  • Contact: Songhua LIU
  • About author:SUO Yifan, born in 1997, M. S. candidate. His research interests include time series anomaly detection, data mining.
    HAO Qiuzhi, born in 2000, M. S. candidate. His research interests include data mining, anomaly detection.
  • Supported by:
    National Natural Science Foundation of China(62176177);Natural Science Foundation of Shanxi Province(202203021211144)

摘要:

在多变量时间序列异常检测任务中,不同变量之间的相关关系复杂,传统的异常检测方法难以明确学习这种相关关系,且多数模型仅考虑变量之间的相关性,对时间依赖性的学习存在不足。因此,提出一种基于高阶特征聚合的时间序列异常检测方法(HFA)。首先,通过图结构学习构造变量之间的关系图;其次,在传统图注意力网络(GAT)的基础上进行改进,充分考虑高阶邻居节点的相关性,更准确地建模变量之间复杂的相关关系;最后,通过融合一维卷积和自注意力机制,充分挖掘序列的时间依赖性。在4个公开数据集上的对比实验结果表明,与次优基线模型Anomaly Transformer相比,HFA的F1分数平均提升了1.34%;与当前主流基线方法TopoGDN(Topology Graph Deviation Network)相比,HFA的F1分数平均提升了9.05%。消融实验结果进一步验证了模型中各个模块的有效性。

关键词: 多变量时间序列, 无监督学习, 图注意力网络, 自注意力, 异常检测

Abstract:

In the anomaly detection tasks for multivariate time series, the correlations between variables are complex, and such correlation is difficult to be learned by traditional anomaly detection methods clearly. In addition, most models only consider the correlation between variables, with learning time dependencies insufficiently. Therefore, a time series anomaly detection method based on High-order Feature Aggregation (HFA) was proposed. Firstly, a variable relationship diagram was constructed through graph structure learning. Secondly, the traditional Graph ATtention network (GAT) was enhanced by taking full account of higher-order neighbor node correlations, thereby modeling complex inter-variable relationships more accurately. Finally, temporal dependencies of the series were captured fully through the integration of one-dimensional convolutions with self-attention mechanism. Experimental results on four public datasets demonstrate that compared with the suboptimal baseline model Anomaly Transformer, HFA has the F1 score increased by 1.34% on average; compared with the current mainstream baseline method TopoGDN (Topology Graph Deviation Network), HFA has the F1 score increased by 9.05% on average. The results of ablation experiments further verify the effectiveness of each module in the model.

Key words: multivariate time series, unsupervised learning, Graph ATtention network (GAT), self-attention, anomaly detection

中图分类号: