Artificial intelligence-based metabolic energy prediction model for animal feed proportioning optimization

Figures & data

Figure 1. Design of Artificial Intelligence metabolic energy prediction model.

Table 1. Nutrient content and variation range of various raw materials.

Raw materials	Metabolic energy	Crude protein	Calcium	Available phosphorous	Lysine	Methionine	Met + Cystine
	(MJ/kg)	(%)	(%)	(%)	(%)	(%)	(%)
Corn	13.56	8.7	0.02 ± 0.003	0.12 ± 0.002	0.24 ± 0.001	0.18 ± 0.002	0.38
Soybean Cake	10.54	40.9	0.30 ± 0.02	0.24 ± 0.009	2.38 ± 0.15	0.59 ± 0.008	1.20
Rice bran cake	10.17	14.7	0.14 ± 0.01	0.22 ± 0.003	0.66 ± 0.04	0.26 ± 0.003	0.56 ± 0.008
Meat and bone meal	8.20	50.0	9.20 ± 0.68	4.70 ± 0.76	2.60 ± 0.053	0.67 ± 0.004	1.00
Lysine	0	0	0	0	98.50	0	0
Methionine	0	0	0	0	0	98.5	98.50
Stone flour	0	0	39.00 ± 0.69	0	0	0	0
Calcium hydrogen phosphate	0	0	20.50 ± 1.08	18.89 ± 0.83	0	0	0
Premix	0	0	0	0	0	0	0
Salt	0	0	0	0	0	0	0

Table 2. Assurance probability, expansion and contraction and weight of each nutritional index.

	Metabolic energy	Crude protein	Calcium	Available phosphorous	Lysine	Methionine	Met + Cystine
	(%)	(%)	(%)	(%)	(%)	(%)	(%)
Confidence coefficient	50	50	69	69	95	95	95
Expansion and contraction	−0.2%	0	0	0	0	−0.2%	0
Weight	5	5	100	5	5	5	5

Table 3. Constraints of nutrients and raw materials.

	Unit	Constraint form	Constrained value
Metabolic energy	MJ/kg	≥	11.92
Crude protein	%	≥	20.00
Calcium	%	≥	0.65
Available phosphorous	%	≥	0.40
Lysine	%	≥	0.90
Methionine	%	≥	0.40
Met + Cystine	%	≥	0.70
Salt	%	≥	0.25
Premix	%	≥	0.35

Figure 2. Performance evaluation results of the Long Short-Term memory (LSTM) model.

Figure 3. Comparison results of Long Short-Term memory (LSTM) and Convolutional Neural Network (CNN) models.

Figure 4. Comparison results of Long Short-Term memory (LSTM) and Recurrent Neural Network (RNN) models.

Figure 5. Training results of the metabolic energy prediction model in the Day time interval.

Figure 6. Training results of the metabolic energy prediction model in the week time interval.

Figure 7. Training results of the metabolic energy prediction model in the Month time interval.

Figure 8. Results of model metabolic energy prediction performance test.

Table 4. Statistics of accuracy results of model prediction.

Data	0.5	1	1.5	2	2.5	3	3.5
Raw data	12.3	11.7	11.5	12.1	11.2	10.9	10.5
Forecast 1	12	11.2	11.1	11.8	11	10.7	10.9
Fitting degree (%)	97.6	95.7	96.5	97.5	98.2	98.2	96.2

Fitting degree was calculated using the simple equation: (forecast 1/raw data) × 100.

Data availability statement

All data generated in the present work is included in the manuscript.