基于生成式对抗网络的自监督多元时间序列异常检测方法
周业瀚, 沈子钰, 周清, 李云

Self⁃supervised multivariate time series anomaly detection based on GAN
Yehan Zhou, Ziyu Shen, Qing Zhou, Yun Li
表3 包含(不包含)自监督模块的CPCGAN模型评价指标情况
Table 3 Performance of CPCGAN with and without self?supervised module
SWaTWADISMDSMAPMSL
PRF1PRF1PRF1PRF1PRF1
CPCGAN (with)0.98150.6610.78990.9910.13160.23230.95110.94840.94970.75810.98220.85570.8820.96860.9232
CPCGAN (without)0.8420.59120.70140.8710.14160.19920.88330.90260.8370.65180.88720.7850.79660.91810.8395