Classification of Parkinson’s disease utilizing multi-edit nearest-neighbor and ensemble learning algorithms with speech samples

Table 5 Classification result in terms of dependence

	Leave-one-out method LOO (%)			Leave-one-subject-out method LOSO (%)
	Accuracy	Sensitivity	Specificity	Accuracy	Sensitivity	Specificity
RF_MENN
Mean	70.93	73.27	67.59	81.50	92.50	70.50
Std	0.031	0.0052	0.0063	0.0105	0.0225	0.0211
Best	76.07	73.65	68.46	88	97	88
RF_MENN_inDe
Mean	55.1	59.2	51.0	65.0	75.0	55
Std	0.003	0.0063	0.0043	0.0211	0.0242	0.0369
Best	55.4	62.9	53.5	67.5	78.5	59.5
SVM_RBF
Mean	63.08	73.08	57.08	67.5	80	55
Std	0	0	0	0	0	0
Best	63.08	73.08	57.08	67.5	80	55
SVM_RBF_inDe
Mean	55.0	62.1	47.8	56.0	71.5	40.5
Std	0	0	0	0	0	0
Best	55.0	62.1	47.8	56.0	71.5	40.5

RF_MENN: reflect the RF + MENN algorithm; RF_MENN_inDe: RF_MENN in the case when a sample is classified and other samples from same subject are not used for building classification model; SVM_RBF: reflect the SVM with RBF kernel; SVM_RBF_inDe: SVM_RBF in the case when a sample is classified and other samples from same subject are not used for building classification model

ISSN: 1475-925X