Anomaly Detection in QAR Data Using VAE-LSTM with Multihead Self-Attention Mechanism

Table 4

Statistical information of four public benchmark datasets.

Dataset	Total length	Train length	Test length	Mean	Std.	Anomaly rate (%)
KPI1	90000	75000	15000	2.3840	0.9174	5.18
KPI2	17562	10000	7562	0.1911	0.1004	0.79
NAB1	18050	15500	2550	37.4794	14.4096	0.08
NAB2	4032	3000	1032	45.1079	1.8774	0.29