Birth weight in Alpine kids using body measurements: a comparison of artificial neural network and multiple linear regression models

Güney, Melis Çelik

doi:10.1590/0103-8478cr20250308

home Tabla de contenido navigate_before Anterior Actual Siguiente navigate_next

Resumen (Inglés)
Resumen (Portugués)

Texto (Inglés)

Download PDF (Inglês)

Resumen (Inglés)
Resumen (Portugués)
Texto (Inglés)
Download PDF (Inglês) (abre em nova aba)

Whatsapp
BlueSky
Mastodon
Facebook
Mais

home Tabla de contenido

Whatsapp
BlueSky
Mastodon
Facebook
Mais

Resumen (Inglés)
Resumen (Portugués)

Texto (Inglés)

Download PDF (Inglês) (abre em nova aba)

ANIMAL PRODUCTION • Cienc. Rural 56 (08) • 2026 • https://doi.org/10.1590/0103-8478cr20250308 linkcopy

Birth weight in Alpine kids using body measurements: a comparison of artificial neural network and multiple linear regression models

Estimativa do peso ao nascer em cabritos alpinos usando medidas corporais: uma comparação entre redes neurais artificiais e modelos de regressão linear múltipla

Authorship SCIMAGO INSTITUTIONS RANKINGS

The present study was conducted to estimate the body weights of Alpine kids using somebody measurements with artificial neural network (ANN) and multiple linear regression (MLR) analysis. For this purpose, the birth weight of 97 kids in total and body measurements such as withers height, rump height, chest depth, chest girth and body length were taken. The model performance criteria used to compare the neural networks and regression analysis results for the goodness of fit are coefficient of determination (R²) and mean square error (MSE). In analyses using artificial neural networks, Levenberg-Marquardt (LM), Bayesian Regularization (BR), and Scaled Conjugate Gradient (SCG) algorithms were applied. The findings showed that the LM algorithm produced the best results and achieved the highest R² and lowest MSE values among the ANN training algorithms. The ANN model (R² = 0.9173; MSE = 0.0006) outperformed the MLR model (R² = 0.791; MSE = 0.101) in predicting the birth weight of Alpine kids.

Key words:
goat production; Türkiye; machine learning; predictive modelling

Authorship

person Melis Çelik Güney *

conceptualization · methodology · writing-review and editing

schoolDepartment of Animal Science, Faculty of Agriculture, Cukurova University, 01380, Adana, Türkiye..Cukurova UniversityTürkiyeAdana, TürkiyeDepartment of Animal Science, Faculty of Agriculture, Cukurova University, 01380, Adana, Türkiye..

0000-0002-6825-6884

E-mail:
celikm@cu.edu.tr.
*Corresponding author

ASSOCIATE EDITOR:

Rudi Weiblen (0000-0002-1737-9817)

SCIENTIFIC EDITOR:

Rudiger Daniel Ollhoff (0000-0002-7972-9472)

DECLARATION OF CONFLICT OF INTEREST

The author declares that there is no conflict of interest.

SCIMAGO INSTITUTIONS RANKINGS

Department of Animal Science, Faculty of Agriculture, Cukurova University, 01380, Adana, Türkiye..Cukurova UniversityTürkiyeAdana, TürkiyeDepartment of Animal Science, Faculty of Agriculture, Cukurova University, 01380, Adana, Türkiye..

Figures | Tables

Figures (2)
Tables (7)

Thumbnail

Figure 1 - Developed ANN model design. WH = withers height; RH = rump height; CD = chest depth; CG = chest girth; BL = body length; BW = body weight; ANN = artificial neural network.

Thumbnail

Figure 2 - The regression fit graphical representation for LM. R = correlation coefficient; LM = levenberg-marquardt algorithm.

Thumbnail

Table 1 Descriptive statistics of birth weight and some body measurements.

Thumbnail

Table 2 The MLR model summaries for prediction of BW.

Thumbnail

Table 3 Results of LM training algorithm of ANN model.

Thumbnail

Table 4 Results of BR training algorithm of ANN model.

Thumbnail

Table 5 Results of SCG training algorithm of ANN model.

Thumbnail

Table 6 The best results for LM, BR and SCG training algorithms.

Thumbnail

Table 7 Comparison of ANN and MLR.

imageFigure 1 - Developed ANN model design. WH = withers height; RH = rump height; CD = chest depth; CG = chest girth; BL = body length; BW = body weight; ANN = artificial neural network. open_in_new

imageFigure 2 - The regression fit graphical representation for LM. R = correlation coefficient; LM = levenberg-marquardt algorithm. open_in_new

table_chartTable 1 Descriptive statistics of birth weight and some body measurements.

Variables	Min	Max	Mean	SE	CV (%)
BW (kg)	2.63	6.12	4.166	0.060	14.275
WH (cm)	33.50	43.00	39.051	0.181	4.577
RH (cm)	33.00	43.50	38.948	0.177	4.498
CD (cm)	10.50	15.50	12.762	0.101	7.796
CG (cm)	29.50	42.00	36.974	0.188	5.012
BL (cm)	26.50	35.50	31.041	0.175	5.554

BW = body weight; WH = withers height; RH = rump height; CD = chest depth; CG = chest girth; BL = body length; Min = minimum value; Max = maximum value; SE = standard error of the mean; CV (%) = coefficient of variation.

table_chartTable 2 The MLR model summaries for prediction of BW.

Variables	Coefficient	Standard Error	T value	Pr ( > \|t\|)
Constant	-6.577	1.034	-6.358	P < 0.001
WH (cm)	0.092	0.092	0.998	0.322
RH (cm)	-0.059	0.088	-0.667	0.508
CD (cm)	-0.048	0.058	-0.824	0.413
CG (cm)	0.149	0.039	3.868	P < 0.001
BL (cm)	0.146	0.031	4.760	P < 0.001

MLR = multiple linear regression; BW = body weight; WH = withers height; RH = rump height; CD = chest depth; CG = chest girth; BL = body length.

table_chartTable 3 Results of LM training algorithm of ANN model.

Transfer functions	Number of neurons	------------------------------------------Performance Criteria-------------------------------------------
		MSE	-----------------------------------------R--------------------------------------------
			Training	Validation	Test	All
Tansig	2	0.00868	0.8540	0.7184	0.8525	0.8364
	4	0.00477	0.8368	0.8377	0.8116	0.8252
	6	0.00337	0.8979	0.8827	0.8411	0.8783
	8	0.00207	0.8950	0.9530	0.7509	0.8523
	10	0.00428	0.8705	0.8652	0.7464	0.8512
	12	0.00296	0.9030	0.6646	0.8202	0.8631
	14	0.00211	0.7604	0.8728	0.8538	0.7834
	16	0.00366	0.7975	0.6553	0.5577	0.7479
	18	0.00426	0.8689	0.8909	0.8526	0.8618
	20	0.00063	0.9102	0.9718	0.9578	0.9333
Logsig	2	0.00766	0.8528	0.8667	0.8308	0.8443
	4	0.00367	0.9048	0.8696	0.7977	0.8691
	6	0.00457	0.8909	0.9325	0.7214	0.8737
	8	0.00276	0.9174	0.5987	0.8817	0.8721
	10	0.00406	0.8542	0.6357	0.8054	0.8290
	12	0.00288	0.8910	0.8313	0.8009	0.8697
	14	0.00194	0.9447	0.6461	0.7830	0.8777
	16	0.00218	0.915	0.7630	0.7148	0.8521
	18	0.00240	0.9056	0.9450	0.8160	0.9034
	20	0.00269	0.8475	0.7192	0.7336	0.8114

LM = levenberg-marquardt algorithm; ANN = artificial neural network; MSE = mean square error; R = correlation coefficient.

table_chartTable 4 Results of BR training algorithm of ANN model.

Transfer functions	Number of neurons	------------------------------------Performance Criteria-------------------------------------
		MSE	------------------------------------R------------------------------------
			Training	Test	All
Tansig	2	0.0073	0.8469	0.8686	0.8513
	4	0.0061	0.8904	0.8202	0.8846
	6	0.0077	0.8462	0.8601	0.8434
	8	0.0078	0.8484	0.8268	0.8430
	10	0.0039	0.8267	0.9133	0.8418
	12	0.0085	0.8442	0.8520	0.8416
	14	0.0076	0.8456	0.8120	0.8422
	16	0.0083	0.8485	0.8307	0.8452
	18	0.0084	0.8250	0.9303	0.8445
	20	0.0066	0.8679	0.7336	0.8445
Logsig	2	0.0055	0.9007	0.6313	0.8599
	4	0.0066	0.8790	0.9266	0.8855
	6	0.0073	0.8535	0.9132	0.8615
	8	0.0058	0.8853	0.8502	0.8776
	10	0.0084	0.8517	0.8476	0.8439
	12	0.0075	0.8477	0.8366	0.8417
	14	0.0067	0.8873	0.5486	0.8697
	16	0.0081	0.8474	0.7993	0.8454
	18	0.0086	0.8319	0.9021	0.8440
	20	0.0085	0.8491	0.8901	0.8454

BR = bayesian regularization algorithm; ANN = artificial neural network; MSE = mean square error; R = correlation coefficient.

table_chartTable 5 Results of SCG training algorithm of ANN model.

Transfer functions	Number of neurons	------------------------------------------Performance Criteria-------------------------------------------
		MSE	-------------------------------------------R-------------------------------------------
			Training	Validation	Test	All
Tansig	2	0.0077	0.8289	0.8474	0.8884	0.8359
	4	0.0110	0.7856	0.9188	0.6362	0.7815
	6	0.0091	0.7817	0.8066	0.7846	0.7812
	8	0.0100	0.5616	0.7799	0.4147	0.5734
	10	0.0100	0.6833	0.8224	0.4298	0.6507
	12	0.0100	0.7789	0.9215	0.7487	0.7902
	14	0.0077	0.7835	0.8874	0.8252	0.8265
	16	0.0074	0.8257	0.7451	0.4940	0.7759
	18	0.0064	0.8699	0.7877	0.7765	0.8410
	20	0.0073	0.8273	0.8789	0.6295	0.7990
Logsig	2	0.0088	0.7989	0.9165	0.7712	0.8066
	4	0.0077	0.7368	0.6455	0.6659	0.7127
	6	0.0075	0.8486	0.8492	0.8941	0.8561
	8	0.0071	0.8616	0.9041	0.8272	0.8601
	10	0.0073	0.7974	0.8176	0.8556	0.8099
	12	0.0076	0.6423	0.5643	0.7236	0.6335
	14	0.0073	0.8750	0.8332	0.7745	0.8599
	16	0.0062	0.8859	0.7726	0.8687	0.8647
	18	0.0090	0.7916	0.7822	0.6779	0.7763
	20	0.0092	0.8107	0.9039	0.7959	0.8192

SCG = scaled conjugate gradient algorithm; ANN = artificial neural network; MSE = mean square error; R = correlation coefficient.

table_chartTable 6 The best results for LM, BR and SCG training algorithms.

Training algorithm	Transfer functions	Number of neurons	MSE	R (Test)
LM	Tansig	20	0.0006	0.9578
BR	Tansig	10	0.0039	0.9133
SCG	Logsig	16	0.0062	0.8687

LM = levenberg-marquardt algorithm; BR = bayesian regularization algorithm; SCG = scaled conjugate gradient algorithm; MSE = mean square error; R = correlation coefficient.

table_chartTable 7 Comparison of ANN and MLR.

Performance criteria	-----------------------------------------------Test data set---------------------------------------------
	MLR	ANN
R²	0.791	0.9173 [R² = (0.9578)²]
MSE	0.101	0.0006

ANN = artificial neural network; MLR = multiple linear regression; R² = coefficient of determination; MSE = mean square error.

How to cite

linkcopy

Ferramentas do artigo

PDFs
Métricas
Figuras e tabelas
Versiones y traducciones
Como citar este artículo
Artículos relacionados

location_on

Universidade Federal de Santa Maria Universidade Federal de Santa Maria, Centro de Ciências Rurais , 97105-900 Santa Maria RS Brazil , Tel.: +55 55 3220-8698 , Fax: +55 55 3220-8695 - Santa Maria - RS - Brazil
E-mail: cienciarural@mail.ufsm.br

rss_feed Acompañe los números de esta revista en su lector de RSS