Parkinson’s Disease Diagnosis in Cepstral Domain Using MFCC and Dimensionality Reduction with SVM Classifier

Table 2

Evaluation of different models in terms of PD detection based on voice data coeff, MFCC coefficients, selected: size of subset of features, Acc (%): percentage of correctly classified subjects, : Mathews correlation coefficient, Sn: sensitivity and Sp: specificity, and HP: hyperparameters

Method	Coefficients	Acc (%)	Sp (%)	Sn (%)	Hyperparameters
LR	7	50.00	60.00	33.33	C = 0.0001
DT	12	51.25	60.00	36.66	d = 10, l = 22
GNB	1	48.12	60.00	28.33	—
LDA	15, 17, 19	62.25	60	66	—
NCC	11	50.00	50	50	—
ANN	11, 17, 18, 20	65.62	85	33	H = 100
SVM (Lin)	10, 15, 16	68.75	90	33.33	C = 10
SVM (Pol)	8	63.75	90	20	degree = 7
SVM (Sig)	14, 15	46.87	70	8.3	C = 0.001
SVM (RBF)	10, 12, 18	77.50	84	73.33	C = 0.01, G = 0.0001