Research Article

A Multiple-Classifier Framework for Parkinson’s Disease Detection Based on Various Vocal Tests

Table 3

Results obtained from applying different methods and classifiers.

ClassifierMethod Accuracy (%)Sensitivity (%)Specificity (%)MCC

-NN (LOSO53.3749.6257.12 0.0007
s-LOO (1–4)42.5030.0055.000.0015
s-LOO (2–5)52.5045.0060.000.0005
s-LOO (3–6)50.0055.0045.000.0000
s-LOO (all)55.0055.0055.000.1000
MCFS67.5075.0060.000.3549
A-MCFS70.0080.0060.000.4082

-NN ()LOSO54.0453.2754.810.0008
s-LOO (1–4)55.0045.0065.000.1021
s-LOO (2–5)60.0055.0065.000.2010
s-LOO (3–6)42.5055.0030.000.0015
s-LOO (all)55.0055.0055.000.1000
MCFS65.0060.0070.000.3015
A-MCFS67.5075.0060.000.3540

-NN ()LOSO54.4253.6555.190.0009
s-LOO (1–4)55.0045.0065.000.1201
s-LOO (2–5)57.5065.0050.000.1517
s-LOO (3–6)50.0070.0030.000.0000
s-LOO (all)55.0070.0040.000.1048
MCFS67.560.0075.000.3540
A-MCFS72.5070.0075.000.4506

-NN ()LOSO53.9454.0453.850.0008
s-LOO (1–4)65.0055.0075.000.3062
s-LOO (2–5)62.5060.0065.000.2503
s-LOO (3–6)42.5065.0020.000.0017
s-LOO (all)57.5065.0050.000.1517
MCFS62.565.0060.000.2503
A-MCFS77.5080.0075.000.5507

SVM (linear kernel)LOSO52.5052.5052.500.0006
s-LOO (1–4)77.5080.0075.000.5507
s-LOO (2–5)70.0080.0060.000.4082
s-LOO (3–6)60.0065.0045.000.2000
s-LOO (all)67.5070.0065.000.3504
MCFS75.0075.0075.000.5000
A-MCFS85.0085.0085.000.6000

SVM (RBF kernel)LOSO55.0060.0050.000.1005
s-LOO (1–4)65.0060.0070.000.3015
s-LOO (2–5)70.0070.0070.000.4000
s-LOO (3–6)72.5070.0075.000.4506
s-LOO (all)65.0070.0060.000.3015
MCFS75.0080.0070.000.5025
A-MCFS87.5090.0085.000.7509

Naïve BayesMCFS75.0090.0060.000.5241
A-MCFS80.0080.0080.000.6000

Discriminant analysisMCFS72.5075.0070.000.4506
A-MCFS82.5080.0085.000.6508

Central tendency metrics used in s-LOO method: 1: mean, 2: median, and 3: trimmed mean (25% removed).
Dispersion metrics used in s-LOO method: 4: standard deviation, 5: mean absolute deviation, and 6: interquartile range.