Data Anomaly Detection through Semisupervised Learning Aided by Customised Data Augmentation Techniques

Table 3

Dataset composition for the SSL training process.

Data pattern	Normal	Missing	Minor	Outlier	Square	Trend	Drift
(a) Validation data composition of each pattern
Quantity	2,208	632	185	87	934	846	108
Percentage (%)	44.16	12.64	3.70	1.74	18.68	16.92	2.16
Total	5,000/100%
(b) Test data composition of each pattern
Quantity	10,689	2,335	1,465	244	2,280	3,712	723
Percentage (%)	49.84	10.89	6.83	1.14	10.63	17.31	3.37
Total	21,448/100%