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

Table 2

The proposed DE-MSVM method performance on axial slice images.

ADNI dataset (axial slice)
Feature selection	Classifier	Accuracy (%)	Sensitivity (%)	Specificity (%)	FOR (%)	FDR (%)	MCC (%)
Without feature selection	Autoencoder	83.4	84.69	84.53	80.51	85.54	70.36
	AdaBoost	82.71	84.99	86.32	75.97	92.76	80.52
	MSVM (linear)	86.25	88.26	72.51	78.45	86.69	74.28
	MSVM	87.7	89.51	87.36	88.93	89.28	87.17
Bat feature selection algorithm	Autoencoder	85.35	85.35	84.56	80.37	86.41	72.06
	AdaBoost	83.15	81.3	82.82	76.04	83.43	80.94
	MSVM (linear)	84.31	83.07	72.11	79.42	82.86	74.63
	MSVM	89.72	90.98	89.8	90.28	89.9	91.63
Grey wolf algorithm	Autoencoder	88.53	87.44	88.09	82.23	88.86	74.46
	AdaBoost	86.33	87.54	88.66	78.58	86.85	81.37
	MSVM (linear)	85.98	85.33	75.05	82.26	89.76	75.99
	MSVM	90	91.85	90.32	91.13	90.59	91.21
Whale optimization algorithm	Autoencoder	87.89	89.75	87.43	84.23	90.34	76.3
	AdaBoost	87.48	86.8	87.47	81.82	91.98	86.78
	MSVM (linear)	89.55	93.26	75.23	82.25	89.17	80.58
	MSVM	95.23	92.76	94.26	95.48	94.6	93.02
Multiobjective differential evolutionary algorithm	Autoencoder	90.98	85	89.86	89.43	90.01	85.61
	AdaBoost	96.67	95.96	95.07	92.82	94.09	91.87
	MSVM (linear)	97.33	96.92	97.41	97.65	96.76	92.35
	MSVM	98.13	98.96	98.2	98.36	97.03	96.06