Table 3.

Accuracy across models and methods

Method	Artificial neural network	Logistic regression	Tree			Linear discriminant	Kernel Naïve Bayes	Support vector machine
			Fine	Medium	Coarse			Linear	Quadratic	Cubic	Fine Gaussian	Medium Gaussian	Coarse Gaussian
CNF1	55.2	62.1	44.8	44.8	44.8	41.4	27.6	51.7	55.2	51.7	48.3	55.2	48.3
CNF2	69	51.7	58.6b	58.6b	58.6b	27.6	37.9	48.3	51.7	51.7	41.4	48.3	41.4
CNF3	31	65.5	34.5	34.5	34.5	31	27.6	44.8	44.8	41.4	41.4	44.8	41.4
DEA	96.3a	77.8a	48.1	48.1	48.1	63	33.3	74.1	74.1	77.8b	40.7	63	59.3

Method	K-nearest neighbors						Ensemble
	Fine	Medium	Coarse	Cosine	Cubic	Weighted	Boosted trees	Bagged trees	Subspace discriminant	Subspace KNN	RUSBoosted trees
CNF1	34.5	44.8	48.3	62.1b	48.3	48.3	48.3	44.8	55.2	44.8	31
CNF2	37.9	41.4	41.4	51.7	37.9	44.8	41.4	44.8	44.8	31	44.8
CNF3	31	37.9	41.4	41.4	34.5	44.8	41.4	27.6	48.3b	27.6	34.5
DEA	70.4	59.3	37	48.1	55.6	70.4	37	55.6a	70.4	40.7	59.3

^aIdentifies the highest accuracy score achieved when implementing ANN, logistic regression, and random forest (codified as bagged trees in MATLAB) to the different categorization methods analyzed

^bIdentifies the highest accuracy score achieved by the DEA hybrid and each alternative configuration through the battery of remaining tests