Table 4.

Percentage classification efficiency for various supervised machine-learning algorithms (linear discriminant analysis, support vector machines, and back propagation neural networking) for breast fillets with woody breast^a and spaghetti meat^b in three different scenarios (all data, without fillet weights, and without fat and protein index value).

Classification method	Subjective classification	Accuracy (all data) c		Accuracy (without fillet weights) d		Accuracy (without fat and protein index) e
		Training (%)	Testing (%)	Training (%)	Testing (%)	Training (%)	Testing (%)
Woody breast meat
Linear discriminant analysis	Normal	72.31	52.63	62.10	43.80	61.70	75.60
	Moderate	43.75	29.41	32.20	17.20	31.30	33.33
	Severe	75.00	59.09	64.00	50.00	68.50	56.30
Support vector machines	Normal	63.86	71.04	60.11	56.52	65.74	73.28
	Moderate	49.88	59.99	50.09	49.77	49.88	50.00
	Severe	71.78	81.48	59.94	64.63	72.49	81.48
Back propagation neural networking	Normal	50.00	47.77	32.85	40.00	42.38	42.22
	Moderate	29.04	23.33	6.67	2.22	7.14	1.12
	Severe	20.95	28.88	14.76	11.12	16.19	14.44
Spaghetti meat
Support vector machines	Normal fillet without spaghetti	69.38	50.00	57.06	60.00	52.15	52.22
	Normal fillet with spaghetti	53.33	50.00	45.5	22.2	50.00	52.35
Back propagation neural networking	Normal fillet without spaghetti	100.00	52.95	65.51	42.30	58.62	50.00
	Normal fillet with spaghetti	100.00	75.00	29.31	19.23	29.31	11.50

Woody breast

^a n = 300 (normal = 148, moderate = 82, severe = 70); spaghetti meat

^b n = 84; All data

^c -WB scores, fillet weight, resistance, reactance, protein index, and fat index; without fillet weights

^d -woody breast (WB) scores, resistance, reactance, protein index, and fat index; and Data without fat and protein index

^e-WB scores, fillet weight, resistance, and reactance.