Modifiable risk factors of vaccine hesitancy: insights from a mixed methods multiple population study combining machine learning and thematic analysis during the COVID-19 pandemic

Table 1 Classification metrics for the best performing model

	Default classification threshold		Optimal classification threshold
		Norwegian sample
AUC		0.94
AUPRC		0.72
Brier score		0.04
F1-score		0.63
Accuracy	94.77%		90.87%
Balanced accuracy	75.83%		86.27%
Recall (sensitivity)	0.53		0.81
Precision (PPV)	0.79		0.48
Specificity	0.96		0.98
		UK sample
AUC		0.98
AUPRC		0.89
Brier score		0.04
F1-score		0.72
Accuracy	94.69%		94.28%
Balanced accuracy	89.09%		89.96%
Recall (sensitivity)	0.81		0.83
Precision (PPV)	0.83		0.79
Specificity	0.96		0.97

ISSN: 1741-7015