Table 2.

Classification performance (sensitivity, specificity, F1) of three different ML models using the NTS data preprocessing pipeline for multiclass prediction of concussion-related impairments. The testing dataset (n = 18) classes were representative of the training set distribution.

Use-Case	Model	Preprocessing Pipeline	Specificity	Sensitivity	F1
Vestibular vs. Neurological vs. Both	RF	NTS	0.93	0.89	0.90
	SVM	NTS	0.83	0.72	0.73
	XGB	NTS	0.93	0.83	0.85