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

Table 3

The proposed DE-MSVM method performance analysis on the sagittal slice.

ADNI dataset (sagittal slice)
Feature selection	Classifier	Accuracy (%)	Sensitivity (%)	Specificity (%)	FOR (%)	FDR (%)	MCC (%)
Without feature selection	Autoencoder	82.7	84.65	82.31	79.87	86.05	79.32
	AdaBoost	82.6	84.96	85.8	75.33	92.19	79.55
	MSVM (linear)	85.47	87.38	83.9	80.89	85.23	79.74
	MSVM	85.73	91.34	89.91	83.14	86.05	84.04
Bat feature selection algorithm	Autoencoder	84.16	88.94	71.49	79.4	86.03	74.09
	AdaBoost	83.82	84.45	86.43	75.15	83.19	80.36
	MSVM (linear)	85.57	85.14	83.94	80.17	85.58	81.44
	MSVM	87.31	89.19	90.93	89.29	89.45	87.87
Grey wolf algorithm	Autoencoder	85.03	90.57	84.6	81.44	88.9	85.6
	AdaBoost	85.41	87.03	88.23	78.21	93.24	80.8
	MSVM (linear)	88.52	86.72	87.36	81.86	88.74	74.46
	MSVM	88.09	91.6	89.8	90.74	90.27	89.84
Whale optimization algorithm	Autoencoder	87.24	89.71	82.98	83.99	84.02	75.5
	AdaBoost	87.26	86.19	82.84	84.38	86.87	86.58
	MSVM (linear)	89.1	90.2	84.39	86.46	88.49	80.54
	MSVM	95.18	94.21	93.27	96.76	89.72	92.19
Multiobjective differential evolutionary algorithm	Autoencoder	97.89	96.78	96.86	97.45	97.05	72.33
	AdaBoost	95.08	95.03	94.6	94.4	94.53	91.42
	MSVM (linear)	90.76	89.07	90.12	90.26	89.2	85.04
	MSVM	98.65	98.32	97.81	98.69	98.78	96.06