Table 4.

Accuracies of 7 classifiers from fivefold cross validation

	ML techniques	PDs vs. Cons (with 36 features)	PDs vs. Cons (with 5 features)	F vs. NF (with 36 features)	F vs. NF (with 6 features)
Accuracy	LR	97.4 ± 2.8	98.0 ± 3.0	72.8 ± 0.8	72.9 ± 10.8
	KNN	94.7 ± 6.9	96.7 ± 4.1	62.9 ± 8.7	61.9 ± 7.8
	NB	91.6 ± 2.9*	96.7 ± 3.3*	70.6 ± 6.5	64.0 ± 9.0
	LDA	96.1 ± 4.3	97.4 ± 3.6	71.8 ± 9.3	69.6 ± 12.3
	QDA	98.0 ± 3.0	97.4 ± 2.8	72.6 ± 11.5	67.5 ± 10.9
	SVM	98.0 ± 3.0	98.0 ± 3.0	69.4 ± 8.9	63.4 ± 16.8
	RF	98.1 ± 1.8	98.0 ± 3.0	79.4 ± 6.9	70.8 ± 10.6

Mean (%) ± standard deviations (%) were calculated through fivefold cross validation; mean values presented in boldface denote the best performance (the highest test accuracy)

ML machine learning, PDs people with PD, Cons controls, F freezers, NF non-freezers, LR logistic regression, KNN k-nearest neighbors, NB Naïve Bayes, LDA linear discriminant analysis, QDA quadratic discriminant analysis, SVM support vector machine, RF random forest

*Denotes a significant difference