Application of Machine Learning in Postural Control Kinematics for the Diagnosis of Alzheimer’s Disease

Table 3

Best results obtained with each classifier trained with datasets with and without the MoCA variable. Between parentheses are the minimum and maximum values obtained. All results are in percentage.

	Accuracy	Sensitivity	Specificity	Best decisional space
SVM with MoCA	91 (75–96.4)	89.3 (64.3–100)	92.7 (71.4–100)	Top 10 features
SVM without MoCA	71,7 (53.6–92.9)	65 (35.7–92.9)	78.4 (35.7–100)
MLP with MoCA	96.6 (96.5–100)	100 (100–100)	94.9 (94.7–100)	Top 11 features and top 15 features
MLP without MoCA	86,1 (79.3–86.2)	78.5 (77.8–78.6)	93.1 (81.8–93.3)
RBN with MoCA	92,5 (75–100)	90.4 (71.4–100)	94.5 (78.6–100)	Top 15 features
RBN without MoCA	74,0 (53.6–82.1)	71.3 (50–100)	76.7 (42.9–100)
DBN with MoCA	96,5 (89.3–100)	95.3 (85.7–100)	97.7 (85.7–100)	Top 15 features
DBN without MoCA	78,0 (57.1–92.9)	79 (14.3–100)	77 (21.4–100)