Differential Evolution and Multiclass Support Vector Machine for Alzheimer’s Classification

Table 4

The proposed DE-MSVM method on the coronal slice.

ADNI dataset (coronal slice)
Feature selection	Classifier	Accuracy (%)	Sensitivity (%)	Specificity (%)	FOR (%)	FDR (%)	MCC (%)
Without feature selection	Autoencoder	77.49	84.79	83.09	75.6	84.42	77.97
	AdaBoost	75.11	82.58	82.41	73.18	84.7	73.39
	MSVM (linear)	79.72	81.07	65.98	73.24	82.65	76.07
	MSVM	85.31	89.77	85.3	86.31	86.2	79.87
Bat feature selection algorithm	Autoencoder	80.07	78.99	77.06	78.33	83.82	78.33
	AdaBoost	77.03	83.98	84.8	70.14	81.35	78.52
	MSVM (linear)	79.75	83.5	79.96	76.71	77.44	70.39
	MSVM	84.1	84.39	85.07	84.8	83.87	80.03
Grey wolf algorithm	Autoencoder	83.29	82.43	81.25	84.96	84.9	79.41
	AdaBoost	86.77	87.32	87.51	80.45	86.66	87.35
	MSVM (linear)	87.09	86.66	93.23	86.98	89.35	84.05
	MSVM	95.07	98.28	90.39	92.28	96.08	89.33
Whale optimization algorithm	Autoencoder	85.41	87.85	84.14	82.68	83.5	80.41
	AdaBoost	90.31	88.03	87.76	89.06	89.82	89.25
	MSVM (linear)	89.62	87.99	86.46	86.59	89.01	88.1
	MSVM	96.21	96.68	95.28	94.71	94.26	95.34
Multiobjective differential evolutionary algorithm	Autoencoder	91.03	90.8	91.34	90.87	90.21	76.59
	AdaBoost	95.9	94.08	94.9	95.49	95.64	81.26
	MSVM (linear)	97.53	96.06	97.65	97.34	96.79	92.7
	MSVM	98.12	97.78	98.7	98.06	97.89	95.81