摘要
针对物联网动态数据流的异常检测需求,提出基于核主成分分析(Principal Component Analysis,PCA)的物联网通信异常数据批量捕获方法。运用可变时间窗口概念构造物联网通信异常数据批量抽取模型,筛选异常数据。结合核方法和PCA方法实现异常数据特征提取,应用孪生神经网络完成异常数据捕获。实验结果表明,运用所提方法得出的捕获结果的F1分数保持在0.85以上,在复杂环境下展现出较强的适应性和可靠性。
To address the anomaly detection requirements for dynamic Internet of Things data streams,a batch capture method for Internet of Things communication anomaly data based on kernel Principal Component Analysis(PCA)is proposed.The concept of a variable time window is employed to construct a batch extraction model for Internet of Things communication anomaly data,enabling the filtering of anomalous data.By combining kernel methods and PCA,the method achieves feature extraction of anomalous data,while utilizing a twin neural network for anomaly detection.Experimental results demonstrate that the F1 score of the capture results remains above 0.85,showcasing stronger adaptability and reliability in complex environments.
作者
董若烟
李文科
缪新萍
吴漾
孔庆波
DONG Ruoyan;LI Wenke;MIU Xinping;WU Yang;KONG Qingbo(Digital Intelligence Operation Center,Guizhou Power Grid Co.,Ltd.,Guiyang 550002,China)
出处
《智能物联技术》
2026年第2期95-98,共4页
Technology of Io T& AI
基金
中国南方电网有限责任公司科技项目(GZKJXM20232513)
贵州省科技计划项目(黔科合支撑[2024]一般055)。
关键词
可变时间窗口
物联网
通信异常数据
数据抽取
variable time window
Internet of Things
communication anomaly data
data extraction
