Table 3.

The best performance of drinking detection using different machine learning models.

Machine Learning Model	Window Size (Samples)	Overlap (%)	Sensitivity (%)	Precision (%)	Specificity (%)	Accuracy (%)
ADA	160	50	86.06	95.50	98.08	94.42
DT	128	50	80.83	95.05	97.91	92.79
RF	96	50	81.87	95.96	98.29	93.34
NB	256	50	92.20	60.28	69.75	76.76
k-NN	128	25	84.87	94.29	97.45	93.68
SVM	224	50	83.17	91.07	96.02	92.14

ADA: AdaBoost; DT: decision tree; RF: random forest; NB: Naïve Bayes; k-NN: k-nearest neighbors; SVM: support vector machine.