Table 3.

Model performance for different chosen augmentations. We can see scaling provides the most benefit for our use case.

Augmentation	Test precision	Test recall	Test F1
No Augmentation	0.9464	0.2755	0.4268
Gaussian Noise	0.9544	0.3116	0.4698
Invert	0.9591	0.3631	0.5267
Scaling	0.8783	0.8442	0.8609
All augmentations	0.9092	0.8618	0.8848