基于无监督对抗学习的时间序列异常检测

邵世宽, 张宏钧, 肖钦锋, 王晶, 刘晓辉, 林友芳

Time series anomaly detection based on unsupervised adversarial learning

Shikuan Shao, Hongjun Zhang, Qinfeng Xiao, Jing Wang, Xiaohui Liu, Youfang Lin

表2 本文的SALAD和基线方法在五个数据集上完全无监督情况下的F1 score

Table 2 F1 score of SALAD and baseline models over five datasets with complete unsupervised settings

模型 A B C D E SALAD 0.7818 0.9373 0.5611 0.6250 0.8094 IF 0.6167 0.0690 0.2718 0.1250 0.4564 AEs 0.1753 0.0823 0.1825 0.0678 0.6117 VAEs 0.1581 0.0552 0.4802 0.0880 0.5613 AAEs 0.3070 0.3600 0.5387 0.2745 0.3268 MAD?GAN 0.3136 0.2391 0.3238 0.1404 0.7467 Beat?GAN 0.1679 0.3417 0.1538 0.2440 0.6268 Donut 0.7674 0.4668 0.5333 0.6000 0.7182