TABLE 4.

Summary of models predicting mean ratings from audio features.

Rating	Model type	R 2	Average R2	Average RMSE
Raspy/Grainy/Rough	RF	0.82	0.78	0.47
	PLSR	0.64	0.56	0.70
Harsh/Noisy	RF	0.56	0.54	0.69
	PLSR	0.43	0.28	0.93
Airy/Breathy	RF	0.45	0.43	0.78
	PLSR	0.36	0.29	0.89
Exertion	RF	0.34	0.32	0.84
	PLSR	0.14	0.09	1.08
Valence	RF	0.68	0.67	0.59
	PLSR	0.56	0.52	0.72

R² values are calculated from models run on the entire dataset (trained using 10-fold cross validation repeated three times). Average R² values are a measure of predictive relevance representing test R² averages across the fivefold cross validation process. RMSE is also averaged across the fivefold.