Table 4.

Comparison of three XGBoost models using features selected in detecting moderate-intensive marijuana intoxication, low-intoxication, and not-intoxicated classes on the test dataset.

Machine learning model	AUCa	F1-score	Recall	Precision	Accuracy
XGBoost-MobiFit	0.99	0.85	0.79	0.92	0.99
XGBoost-Mobile	0.96	0.72	0.75	0.70	0.97
XGBoost-Fitbit	0.97	0.74	0.70	0.79	0.98

^aAUC: area under the curve.