基于时空图注意力网络的多变量时序数据异常检测方法

肖刚,卢大鹏,郑文博,程振波,张元鸣

Multivariable time series data anomaly detection method based on spatiotemporal graph attention network

Gang XIAO,Dapeng LU,Wenbo ZHENG,Zhenbo CHENG,Yuanming ZHANG

表 3 所提模型的模块消融实验结果

Tab.3 Modular ablation experimental results for proposed model %

模型 SWaT MSL SMD PSM P R F1 P R F1 P R F1 P R F1 TNO 85.59 88.29 86.91 76.42 85.61 80.75 83.35 80.26 84.25 93.71 89.82 91.72 GCNO 72.15 78.27 75.08 74.45 76.50 75.46 60.55 66.74 63.49 78.93 82.64 80.74 STGATA 86.78 88.50 87.63 85.71 81.54 83.57 87.36 84.07 85.68 90.02 92.93 91.45 STGATC 85.13 81.08 83.05 84.41 82.51 84.42 84.90 82.71 83.79 88.45 85.42 86.90 STGAT 90.99 95.37 93.13 89.74 86.85 88.06 89.95 85.04 87.42 98.49 96.62 97.54