Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice BMC Vet Res . 2026 Mar 5;22:218. doi: 10.1186/s12917-026-05349-3 Search in PMC Search in PubMed View in NLM Catalog Add to search Application of XGBoost and Random Forest algorithms for body weight prediction in Blackbelly sheep using biometric measurements Miguel Ángel Gastelum Delgado Miguel Ángel Gastelum Delgado 1 Division Académica de Ciencias Agropecuarias, Universidad Juárez Autónoma de Tabasco, Carr. Villahermosa-Teapa, km 25, Villahermosa, Tabasco, CP 86280 México Find articles by Miguel Ángel Gastelum Delgado 1 , Cem Tırınk Cem Tırınk 2 Department of Animal Science, Faculty of Agriculture, Igdir University, TR76000, Iğdır, Türkiye Find articles by Cem Tırınk 2 , Rosa Inés Parra-Cortés Rosa Inés Parra-Cortés 3 Universidad de Ciencias Aplicadas y Ambientales U.D.C.A, Área de Ciencias Agropecuarias, Grupo de Investigación en Ciencia Animal, Bogotá, CP 111166 Colombia Find articles by Rosa Inés Parra-Cortés 3 , Ignacio Vázquez Martínez Ignacio Vázquez Martínez 1 Division Académica de Ciencias Agropecuarias, Universidad Juárez Autónoma de Tabasco, Carr. Villahermosa-Teapa, km 25, Villahermosa, Tabasco, CP 86280 México Find articles by Ignacio Vázquez Martínez 1 , Armando Gomez-Vazquez Armando Gomez-Vazquez 1 Division Académica de Ciencias Agropecuarias, Universidad Juárez Autónoma de Tabasco, Carr. Villahermosa-Teapa, km 25, Villahermosa, Tabasco, CP 86280 México Find articles by Armando Gomez-Vazquez 1 , Aldenamar Cruz-Hernandez Aldenamar Cruz-Hernandez 1 Division Académica de Ciencias Agropecuarias, Universidad Juárez Autónoma de Tabasco, Carr. Villahermosa-Teapa, km 25, Villahermosa, Tabasco, CP 86280 México Find articles by Aldenamar Cruz-Hernandez 1 , Enrique Camacho-Pérez Enrique Camacho-Pérez 4 Facultad de Ingeniería, Universidad Autónoma de Yucatán, Av. Industrias No Contaminantes s/n, Mérida, Yucatán México Find articles by Enrique Camacho-Pérez 4 , Dany Alejandro Dzib-Cauich Dany Alejandro Dzib-Cauich 5 Tecnológico Nacional de México, Instituto Tecnológico Superior de Calkiní, Av. Ah-Canul, Calkiní, Campeche, C.P. 24900 México Find articles by Dany Alejandro Dzib-Cauich 5 , Hasan Önder Hasan Önder 6 Department of Animal Science, Faculty of Agriculture, Ondokuz Mayis University, TR55139, Samsun, Türkiye Find articles by Hasan Önder 6 , Uğur Şen Uğur Şen 7 Department of Agricultural Biotechnology, Faculty of Agriculture, Ondokuz Mayis University, TR55139, Samsun, Türkiye Find articles by Uğur Şen 7 , Kadyrbai Chekirov Kadyrbai Chekirov 8 Department of Biology, Faculty of Sciences, Kyrgyz Turkish Manas University, Bishkek, 720044 Kyrgyzstan Find articles by Kadyrbai Chekirov 8 , Yüksel Aksoy Yüksel Aksoy 9 Department of Animal Science, Faculty of Agriculture, Eskişehir Osmangazi University, TR26160, Eskişehir, Türkiye Find articles by Yüksel Aksoy 9 , Hilal Tozlu Çelik Hilal Tozlu Çelik 10 Department of Food Processing, Vocational School of Ulubey, Ordu University, TR52850, Ulubey, Ordu, Türkiye Find articles by Hilal Tozlu Çelik 10 , Olatunbosun Odu Olatunbosun Odu 11 Department of Animal Science, Faculty of Agriculture, University of Ibadan, Ibadan, Nigeria Find articles by Olatunbosun Odu 11, ✉ , Alfonso J Chay-Canul Alfonso J Chay-Canul 1 Division Académica de Ciencias Agropecuarias, Universidad Juárez Autónoma de Tabasco, Carr. Villahermosa-Teapa, km 25, Villahermosa, Tabasco, CP 86280 México Find articles by Alfonso J Chay-Canul 1 Author information Article notes Copyright and License information 1 Division Académica de Ciencias Agropecuarias, Universidad Juárez Autónoma de Tabasco, Carr. Villahermosa-Teapa, km 25, Villahermosa, Tabasco, CP 86280 México 2 Department of Animal Science, Faculty of Agriculture, Igdir University, TR76000, Iğdır, Türkiye 3 Universidad de Ciencias Aplicadas y Ambientales U.D.C.A, Área de Ciencias Agropecuarias, Grupo de Investigación en Ciencia Animal, Bogotá, CP 111166 Colombia 4 Facultad de Ingeniería, Universidad Autónoma de Yucatán, Av. Industrias No Contaminantes s/n, Mérida, Yucatán México 5 Tecnológico Nacional de México, Instituto Tecnológico Superior de Calkiní, Av. Ah-Canul, Calkiní, Campeche, C.P. 24900 México 6 Department of Animal Science, Faculty of Agriculture, Ondokuz Mayis University, TR55139, Samsun, Türkiye 7 Department of Agricultural Biotechnology, Faculty of Agriculture, Ondokuz Mayis University, TR55139, Samsun, Türkiye 8 Department of Biology, Faculty of Sciences, Kyrgyz Turkish Manas University, Bishkek, 720044 Kyrgyzstan 9 Department of Animal Science, Faculty of Agriculture, Eskişehir Osmangazi University, TR26160, Eskişehir, Türkiye 10 Department of Food Processing, Vocational School of Ulubey, Ordu University, TR52850, Ulubey, Ordu, Türkiye 11 Department of Animal Science, Faculty of Agriculture, University of Ibadan, Ibadan, Nigeria ✉ Corresponding author. Received 2025 Oct 2; Accepted 2026 Jan 30; Collection date 2026. © The Author(s) 2026, modified publication 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13072480 PMID: 41787524 This article has been corrected. See BMC Vet Res. 2026 Jul 30;22:451 . Abstract Background Reliable and cost-effective live weight determination is critical for improving both management and selection efficiency in small-scale sheep farming. In recent years, machine learning approaches based on phenotypic biometric measurements have emerged as a practical alternative for producers with limited access to weighing infrastructure. This study aimed to compare the performance of two machine learning algorithms (XGBoost and Random Forest) for estimating live weight based on biometric measurements in Blackbelly sheep (60 females, 60 males) raised in humid tropical climates. Results Using the obtained biometric measurements, both algorithms were evaluated separately for training and test sets. The XGBoost algorithm demonstrated high accuracy on the training data (R 2 = 0.981; MSE = 0.416; RMSE = 0.645; MAE = 0.486; AIC = 150.513 and BIC = 157.841) but had limited generalization to the test data (R 2 = 0.813; MSE = 3.296; RMSE = 1.816; MAE = 1.440; AIC = 146.125 and BIC = 150.791). In contrast, the Random Forest algorithm produced more balanced and stable predictions on both training (R 2 = 0.969; MSE = 0.630; RMSE = 0.794; MAE = 0.617; AIC = -31.230 and BIC = -21.459) and test data (R 2 = 0.873; MSE = 2.188; RMSE = 1.479; MAE = 1.105; AIC = 35.401 and BIC = 41.623). Optimal hyperparameters were determined for both models, and model fit criteria were compared. The results revealed that Random Forest offers a more reliable and stable option for estimating live weight. Conclusions The current study demonstrates that machine learning models based on phenotypic biometric measurements can be used in decision-support processes in small-scale sheep farming operations. The Random Forest algorithm stands out as a suitable tool for improving production and selection efficiency in live weight estimation. Furthermore, validation studies encompassing different sheep breeds and larger data sets will strengthen the reliability and generalizability of the developed models. Keywords: Random forest, XGBoost, Blackbelly, Sheep, Machine learning Background Sheep have played a significant role in obtaining animal-based products (meat, milk, and wool) for many civilizations from the past to the present and maintain their importance as multipurpose small ruminants that contribute to developing rural economies within the framework of sustainable development goals [ 1 ]. In this context, sustainable animal breeding practices are critical in preserving the ecological balance and efficient use of natural resources [ 2 ]. In addition, using highly adaptable animals for each region within the scope of sustainable animal breeding is extremely important. In this context, Blackbelly sheep, which are perfectly adapted to Mexico’s tropical and subtropical climate conditions, stand out as a species that supports sustainable livestock practices in the region. In particular, the resilience of these sheep and their efficient reproductive capacity, even in conditions of underdeveloped infrastructure, contribute to the development of the Mexican rural economy and play a major role in strengthening the sources of income for local communities [ 3 ]. The hair sheep breeds are the main genotypes used in sheep production systems in the tropics [ 4 ]. Historically, these production systems used Pelibuey and Blackbelly as maternal breeds. This has led to the development of production systems based on the crossbreeding of flocks of Pelibuey, Blackbelly and Katahdin sheep and their pure-bred progeny [ 5 ]. In these systems, the continuous determination of animal growth is a major challenge for smallholder farmers due to several factors [ 6 ]. Blackbelly sheep are a genetic resource that is perfectly adapted to the humid tropical climates of Mexico and can be considered as a valuable option to improve sheep breeding in the humid tropics of the world. The characterization of carcass tissue composition is essential to promote the economic efficiency of these production systems [ 7 , 8 ]. In addition to all the sustainable farming practices and genetic adaptation efforts mentioned above, the economic returns obtained from meat animals are based on body weight. Because the incomes obtained by breeders are directly based on the weight of the animals and this weight plays a decisive role in marketing and commercial evaluation processes [ 9 ]. In this context, accurate determination of animal body weight is of critical importance in order to maximize meat production. Therefore, studies conducted to explain the relationship between biometric measurements and animal weights are of increasing interest both in terms of scientific research and applied animal husbandry, and the development of methodologies in this field has an important place in achieving sustainable animal breeding aims [ 2 ]. In this context, body weight (BW) plays an important role in decisions regarding herd management as one of the most accurate methods of determining growth [ 10 , 11 ]. The traditional weighing method has a negative impact as it involves transporting sheep to a weighing station, increasing labor, and creating stressful conditions for animals [ 11 , 12 ]. Additionally, this can result in a live weight loss of 1.8–2.9 kg or 3.5–5.6% in small ruminants, such as sheep [ 11 , 12 ]. The lack of portable weighing platforms or their unavailability in small livestock farms makes direct measurement of live weight difficult. However, basic biometric measurements such as heart circumference, withers height, and body length can be easily obtained with simple equipment and provide reliable indicators for estimating live weight. Furthermore, various techniques are used to identify the body weight of livestock, largely due to technological advances in both hardware and computational algorithms, instead of classical weighing approaches [ 13 ]. The identification of BW as a more valuable tool for managing sheep production, monitoring growth and performance, and assisting producers in making decisions such as optimal feed amount, medicinal doses, marketing price, and optimum slaughtering time per sheep about sustainable livestock farming [ 2 ]. It is thought that using both hardware and computational algorithms, rather than classical weighing approaches, will eliminate stress factors that occur especially during the weighing of animals, thereby improving animal welfare. There are many studies in the literature on estimating live body weight in various species such as dog, cattle, rabbit, sheep and camel based on biometric measurements. In these studies, various statistical techniques such as multiple linear regression, decision trees, Multivariate Adaptive Regression Splines (MARS), and Artificial Neural Networks (ANNs), etc., were used in different species and different breeds within species [ 14 – 17 ]. In recent studies on sheep, the subject of our current study, in addition to traditional statistical methods, there has been a trend towards more advanced and higher generalization ability machine learning techniques, especially regression-based and decision tree-based machine learning algorithms. This trend has been observed in recent years, especially due to its capacity to analyze heterogeneous data structures and large data sets more effectively, further increasing the importance of machine learning techniques in scientific research and industrial applications. These techniques play a critical role in optimizing sheep farming practices by providing more precise and predictive analyses compared to traditional methods. In this context, especially when the basic assumptions of multiple linear regression analysis are not provided, modern statistical approaches that do not require distribution assumptions provide much more understandable and reliable results. The use of XGBoost and Random Forest algorithms in estimating the body weight of Blackbelly sheep breed is limited in the literature. Therefore, evaluating these algorithms specifically for this breed is considered to be a valuable contribution to breed-specific studies. This provides an opportunity to evaluate the high predictive performance of the algorithms and their ability to effectively model complex data structures. It is thought that this study will contribute significantly to the existing literature in the field by revealing the potential applications and efficiency of these algorithms. In most of the studies to date, multiple linear regression analysis has been used to predict the body weight of animals. However, the use of machine learning methods to accurately predict body weight from biometric measurements appears promising, as the relationship between body weight and biometric measurements is non-linear [ 18 ]. In the current study, it is aimed to compare the performances of two machine learning algorithms (XGBoost and Random Forest) for estimating live weight based on biometric measurements in Blackbelly sheep. Materials and methods Animals and experimental area In this study, data from 120 Blackbelly lambs (60 females and 60 males) were analyzed. The animals were obtained from a local farm specializing in the breeding of this genotype. The experiment was conducted at the Sureste Ovine Integration Center (CIOS, 17°78’N, 92°96’W”; 10 masl). It is in the R/a Alvarado Santa Irene 2da Secc, Centro municipality, Tabasco, Mexico. The area has a humid tropical climate. Temperatures range from 15°C to 44°C, with an average of 26°C. Each measurement, including body weight (BW), heart circumference (HG), cross-body length (DBL), abdominal circumference (AG), body length (BL), withers height (WH), rump height (RH), and hip-width (HW), was recorded, considering animal welfare conditions. The lambs were clinically healthy and were between 6 and 8 months of age. BW was recorded using a fixed platform scale with a capacity of 300 kg and an accuracy of 20 g, while the biometric measurements were taken using a flexible fiberglass tape measure (Truper ® ) as described by [ 19 ]. Statistical analysis An independent-samples t-test was used to assess the effect of sex on the biometric characteristics examined. The level of significance was accepted as p < 0.05. The XGBoost algorithm, which has been continuously improved by many scientists over time, was proposed by Chen and Guestrin in 2016 [ 20 ]. XGBoost helps to solve many data science problems, such as prediction and classification problems, by providing fast and accurate parallel tree boosting [ 21 ]. The XGBoost algorithm, an advanced implementation of the gradient boosting algorithm, is designed to be highly efficient, flexible and portable with optimized structure [ 22 ]. XGBoost works by creating a collection of decision trees, each trained on a different subset of the data, and the resulting trees are then combined to minimize the error associated with the prediction. By combining the results of several trees, XGBoost can make more accurate predictions than a single decision tree [ 23 ]. Unlike other boosting algorithms, the XGBoost algorithm also allows for regularizations that help prevent overfitting [ 24 ]. In this way the algorithm stands out for its ability to perform quickly and effectively, especially on large datasets, in addition to providing adjustments that help prevent overfitting. The algorithm minimizes the error rate while also considering the model’s complexity, which allows for balancing the overall model performance. In particular, tuning the hyperparameters of XGBoost allows the model to strike an appropriate balance between bias and variance. These include hyperparameters such as maximum depth, eta, etc [ 22 ]. In the current study, two primary hyperparameters, eta and max_depth, were used for the XGBoost algorithm. The eta hyperparameter determines the model’s learning rate; smaller values result in slower learning and can reduce the risk of overfitting. Keeping this parameter low allows the model to generate more trees, resulting in a more balanced and generalizable structure. Max_depth represents the maximum depth of each decision tree; higher values allow for learning more complex structures but can also lead to overfitting [ 20 ]. Understanding the impact of each parameter on the model is critical when determining the optimal parameter combination. Therefore, XGBoost continues to be a preferred method for obtaining high-performance and reliable results in machine learning applications, especially when dealing with complex data structures. Considering this information, the XGBoost can be regarded as an excellent choice for predicting sheep body weight. The Random Forest algorithm (RFR) is a standard procedure among multivariate statistical methods, particularly useful for solving regression and classification problems due to its practicality. The RFR algorithm consists of a process that adds a layer of randomness to the bagging algorithm. The RFR algorithm, proposed by Breiman [ 25 ], combines sets of regression trees hierarchically from the root to the leaf, utilizing a set of constraints [ 26 , 27 ]. The most important advantage of this algorithm is that it can be easily used in nonlinear situations in solving classification and prediction-type problems [ 28 ]. The algorithm has three phases, and the first phase is assembling several trees (ntree) from the original data. The second phase is to create an untrimmed regression or classification tree for each sample. The final phase involves predicting the final data of the tree obtained from the algorithm [ 29 ]. The dataset was randomly divided into two groups: 70% for training and 30% for testing. During the modeling process, 10-fold cross-validation was applied separately for each algorithm to evaluate their generalization performance. The hyperparameter optimization process was carried out using the grid search method which was performed for the XGBoost algorithm using the parameters ‘eta’ (range 0.01–0.3, increments of 0.01) and ‘max_depth’ (range 3–10, increments of 1) and the Random Forest algorithm using the parameters ‘mtry’ (number of variables starting from 2 to 1, increments of 1) and ‘nodesize’ (range 1–10, increments of 1). These analyses were conducted using a systematic and highly sensitive screening approach to ensure that the algorithms reached optimal performance levels. To assess model fit, the dataset was divided into training and test sets. Each algorithm was trained using only the training data and then independently evaluated on the test data. Model performance was measured using the coefficient of determination (R 2 ), mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). Reliable algorithms can be determined by using the highest R 2 and the lowest values of MSE, RMSE, MAE, AIC, and BIC. These metrics were calculated separately for both the training and test sets to assess the model’s applicability and generalizability. All statistical analyses were performed using R and Python software [ 30 , 31 ]. Descriptive statistical methods were preferred to establish the necessary definitions for the data sets. Relevant statistics for explanatory and dependent variables were obtained with the “psych” package of the R software [ 32 ]. Pearson correlation analysis was performed to visualize the relationship between explanatory and dependent variables, and the “corrplot” package in R was used during this analysis [ 33 ]. The separation of the data set into training and test sets was performed using the “caret” package [ 34 ]. The “randomForest” and “xgboost” packages were preferred for applying the XGBoost and Random Forest algorithms [ 29 , 35 ]. Three-dimensional surface plots were visualized using the Python software. Results In Table 1 , descriptive statistics of explanatory and response variables are presented separately for female and male sheep. In addition, significant sex effects were observed on some biometric characteristics from the results of an independent-samples t-test. Significant differences were found between males and females, particularly in body weight and heart circumference ( p < 0.05). Table 1. Descriptive statistics of explanatory and response variables within the scope of sex factor Variables Sex Mean ± Standard Error Min-Max Coefficient of Variation (%) Sig. BW Female 26.64 ± 0.51 19.55–36.45 14.87 0.000 Male 30.71 ± 0.40 23.20–37.00 10.31 HG Female 69.27 ± 0.54 60.00–77.00 6.13 0.000 Male 73.62 ± 0.55 63.00–81.00 5.85 DBL Female 49.75 ± 0.38 44.00–57.00 6.05 0.141 Male 49.08 ± 0.47 40.00–56.00 7.53 AG Female 75.32 ± 0.72 65.00–88.00 7.45 0.073 Male 76.90 ± 0.62 65.00–88.00 6.32 BL Female 45.00 ± 0.32 40.00–51.00 5.62 0.000 Male 42.40 ± 0.53 34.00–53.00 9.79 WH Female 64.30 ± 0.49 53.00–74.00 5.99 0.196 Male 63.73 ± 0.43 56.00–72.00 5.25 RH Female 63.13 ± 0.44 56.00–72.00 5.44 0.018 Male 64.50 ± 0.46 58.00–77.00 5.61 HW Female 14.54 ± 0.17 11.50–18.20 9.55 0.004 Male 13.88 ± 0.16 11.00–18.00 9.43 Open in a new tab BW body weight, HG heart circumference, DBL diagonal body length, AG abdominal circumference, BL body length, WH withers height, RH rump height, HW hip-width Sig: NS = p ≥0.05 (not significant), * = p < 0.05, ** = p < 0.01, *** = p < 0.001 The mean body weight (BW) for female sheep was 26.64 ± 0.51 kg, while in male sheep it is 30.71 ± 0.40 kg. The coefficient of variation for BW is 14.87%. The coefficient of variation (CV(%)) was calculated to be 14.87% for BW in females, while in males, the CV(%) was calculated to be 10.31% for body weight. In both sexes, CV (%) rates for HG, WH, and RH variables were below 6%, demonstrating low variation. Figure 1 presents the distribution characteristics of explanatory and response variables within the sex factor using box plots. Mean values for variables such as BW, HG, AG, and RH were observed to be higher in males than in females. Means for variables such as BL and HW were slightly higher in females than in males. Error bars indicate the distribution ranges of the variables measured in both sexes and are consistent with the coefficients of variation (as shown in Table 1 ). This visualization allows for the identification of differences in the distribution of biometrical traits by sex. Fig. 1. Open in a new tab Evaluation of biometrical characteristics according to sex with boxplot and error bars Figure 2 shows the Pearson correlation coefficients for biometrical variables by sex. In female Blackbelly sheep, there is a high positive correlation (0.89) between BW and HG, and a significant positive correlation (0.83) between BW and WH. Additionally, a significant correlation (0.73) is found between HW and BW. In males, the correlation between BW and HG is 0.85, and between BW and AG is 0.75. A positive correlation (0.72) is also observed between BW and HW. Fig. 2. Open in a new tab Correlation analysis results When all individuals are evaluated together, the highest correlation coefficient is between BW and HG, at 0.90. Similarly, the correlations between BW and WH (0.76) and HW (0.49) are also significant. Some sex-related differences in the correlation patterns are noted. For example, the BL (body length) variable shows a higher correlation with BW in males (0.56), while the relationship is weaker in females (0.19). Figures 3 , 4 and 5 , and 6 present detailed performance metrics of the XGBoost and Random Forest algorithms on the training and test datasets, reflecting the effects of the hyperparameter optimization process on model performance. Each graph visually demonstrates both the success of the respective algorithm during the training process and its generalizability to external examples. Fig. 3. Open in a new tab Surface plot for XGBoost algorithm results within the scope of goodness-of-fit criteria (train set) Fig. 4. Open in a new tab Surface plot for XGBoost algorithm results within the scope of goodness-of-fit criteria (test set) Fig. 5. Open in a new tab Surface plot for Random Forest algorithm results within the scope of goodness-of-fit criteria (train set) Fig. 6. Open in a new tab Surface plot for Random Forest algorithm results within the scope of goodness-of-fit criteria (test set) Figures 3 and 4 show the performance of the XGBoost algorithm in the training and test datasets, respectively. In the training dataset, the XGBoost algorithm produced low MSE, RMSE, and MAE values along with a high R 2 value (0.98). These results indicate that the model has a high fit on the training data. In the test dataset, a decrease in the R 2 value was observed, and an increase in error metrics occurred. This reveals the performance difference between the training and test datasets. Figures 5 and 6 present the performance results of the Random Forest algorithm in the training and test datasets. The Random Forest algorithm obtained a high R 2 value (0.97) in the training dataset, while the R 2 value was 0.87 in the test dataset. The MSE, RMSE, and MAE values calculated in the test dataset were lower compared to the XGBoost algorithm. Furthermore, the AIC and BIC values obtained in the test dataset were also found to be lower for the Random Forest algorithm. When the results presented in Figs. 3 , 4 , 5 and 6 are evaluated together, it is seen that both algorithms exhibit different performance profiles in the training and test datasets. These findings allow for a comparison of the algorithms in terms of prediction accuracy and error metrics. Table 2 presents the optimum hyperparameters and model fit criteria for eXtreme Gradient Boosting (XGBoost) and Random Forest. Among the most suitable hyperparameters for the XGBoost algorithm, the maximum depth (max_depth) was determined as five, and the learning rate (eta) as 0.05. In the Random Forest algorithm, the best performance was obtained with an mtry value of 4 and a node size value of 1. In addition to this information, although no major changes were determined for max_depth for the XGBoost algorithm in Figs. 3 , 4 , 5 and 6 , it was observed that there were major differences in eta values. However, as a result of the examinations made for both node size and mtry hyperparameters in the Random Forest algorithm, it was seen that a more stable model was obtained compared to the hyperparameter changes in XGBoost. Table 2. The optimum hyperparameters and the results of goodness-of-fit criteria Optimal hyperparameters of each algorithm XGBoost Random Forest max_depth 5 mtry 4 eta 0.05 nodesize 1 Goodness-of-fit criteria XGBoost Training Testing Random Forest Training Testing R 2 0.981 0.813 R 2 0.969 0.873 MSE 0.416 3.296 MSE 0.630 2.188 RMSE 0.645 1.816 RMSE 0.794 1.479 MAE 0.486 1.440 MAE 0.617 1.105 AIC 150.513 146.125 AIC -31.230 35.401 BIC 157.841 150.791 BIC -21.459 41.623 Open in a new tab R 2 coefficient of determination, MSE mean square error, RMSE root mean square error, MAE mean absolute error, AIC Akaike’s information criterion, BIC Bayesian information criterion According to the model fit criteria, the performance metrics of both algorithms were evaluated on the training and test sets. While a significantly high R-squared (R 2 = 0.981) value was obtained for the training set using the XGBoost algorithm, this value was lower for the test set (R 2 = 0.813). A similar situation was observed in the Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) values. While the performance on the training set was quite good, a decrease in performance on the test set was observed. On the other hand, the Random Forest algorithm presented a higher R-squared value (R 2 = 0.873) compared to XGBoost on the test set. In addition, the MSE, RMSE and MAE values of the Random Forest algorithm were lower than XGBoost on the test set, indicating that the model generally has a better generalization ability. Model performance was also evaluated using AIC and BIC metrics. For the XGBoost algorithm in the test dataset, the AIC and BIC values were calculated as 146.125 and 150.791, respectively. For the Random Forest algorithm, the AIC and BIC values in the test dataset were obtained as 35.401 and 41.623, respectively. These values allow for a quantitative comparison of the model fit and complexity of both algorithms on the test dataset. Figure 7 shows the variable significance levels (relative significance scores showing the contribution of each variable to live weight estimation) calculated for the XGBoost and Random Forest algorithms. In both algorithms, the chest circumference (HC) variable was determined to have the highest significance level. The significance level of the HC variable was calculated as 59.91% in the XGBoost algorithm and 49.42% in the Random Forest algorithm. Fig. 7. Open in a new tab Variable importance for each algorithm The AG variable ranks second with a significance level of 21.38% in the XGBoost algorithm and 23.15% in the Random Forest algorithm. The significance level of the sex variable was determined as 7.55% in the Random Forest algorithm and 5.36% in the XGBoost algorithm. It was observed that the significance levels of other variables, such as hip width (HW) and rump height (RH), differed between the algorithms. The results presented in Fig. 7 quantitatively demonstrate the relative importance levels of the variables used by both algorithms in live weight estimation. This graph allows us to identify which variables the models give more weight to in the estimation process and objectively visualizes how the relative effects of the variables on live weight differ according to the algorithms. These comparative importance levels of the XGBoost and Random Forest algorithms provide an explanatory framework for the decision-making mechanisms of the algorithms and reveal, with empirical data, which inputs the models are more sensitive to. In this way, it becomes possible to understand the differences between the algorithms in variable selection at a numerical level. Discussion This study compared the performance of the XGBoost and Random Forest algorithms for estimating live weight in Blackbelly sheep. Findings obtained on the training and test datasets indicate that the algorithms have different learning dynamics. While XGBoost demonstrated high accuracy on the training set, it showed more limited generalization performance on the test set. In contrast, the Random Forest algorithm was found to offer more balanced and consistent performance on both datasets. These findings suggest that XGBoost may be more susceptible to overfitting due to its greater capacity to learn within-sample variance, while Random Forest can produce lower-variance estimates thanks to bootstrap sampling and variable randomization [ 36 ]. In the present study, the performance of XGBoost and Random Forest algorithms in estimating live weight from biometric measurements was examined, and the behavior of both algorithms in training and test sets was systematically compared. The findings in this context are consistent with studies conducted with similar data structures in the literature. For example, in a study conducted on Corriedale sheep by Canazo-Cayo et al. [ 37 ], it was reported that the Random Forest algorithm gave superior results compared to other methods in performance metrics such as R 2 , RMSE, and MAE in models created using biometric measurements; this is parallel to the relatively higher generalization ability of Random Forest in the test set in the present study. Alsahaf et al. [ 38 ] compared the XGBoost and Random Forest algorithms for slaughter weight estimation in pigs and reported that both algorithms achieved high accuracy. However, it is natural that these results differ from those of the current study, especially considering the smaller sample size. Similarly, Hamadani and Ganai [ 39 ] compared numerous algorithms for estimating live weight in sheep and demonstrated that chest circumference and body length are crucial factors in determining model performance. In addition, Faraz et al. [ 22 ] stated that XGBoost is an effective and reliable method for estimating live weight in a study conducted on the Kajli sheep breed comparing XGBoost and MARS algorithms. This finding is consistent with the high performance of XGBoost in explaining a large portion of the training data in the present study. However, the decrease in the generalization performance of this algorithm in the test set has also been reported in other studies in the literature, pointing to the effectiveness of factors such as data structure, sample size, and feature distribution that lead to performance variability among different algorithms [ 40 ]. In this study, linear measures such as HG, WH, and BL were identified as the most influential variables in the variable importance results. These findings are highly biologically consistent. HG is directly related to lung capacity, rib cage volume, skeletal development, and muscle mass, and is one of the strongest determinants of body weight in ruminants [ 41 – 43 ]. Furthermore, a strong positive relationship between body length, withers height, and live weight has been reported in many studies [ 44 , 45 ]. Since these anatomical measures reflect the animal’s skeletal size and muscle tissue capacity, it is natural and expected that the algorithms would assign high importance to these variables [ 46 ]. The performance differences observed between the algorithms may largely be due to sample size, variable distribution, breed-specific phenotypic characteristics, and hyperparameter settings. For example, the higher performance of XGBoost in some studies (e.g., Coşkun et al. [ 47 ]; Esener and Eşki [ 48 ]) can be explained by the model’s sensitivity to sample size. Random Forest’s higher generalization performance on the test set is related to its more stable learning strategy in heterogeneous data structures [ 25 ]. In this context, this study, which compares the XGBoost and Random Forest algorithms in the Blackbelly sheep breed, expands a research area that has received limited coverage in the literature and presents findings on body weight estimation for this breed. The study’s strengths include the compatibility of variable importance ranking with biological underpinnings, a critical analysis of methodological differences, and literature comparisons. Future studies, which conduct similar modeling with larger sample sizes, different breeds, and under various environmental conditions, will provide the opportunity to more comprehensively evaluate the generalizability and biological validity of the algorithms. In addition to these original contributions, the study’s findings have implications for practical applications that are also noteworthy. The findings demonstrate the development of machine learning-based decision support systems that provide fast, economical, and practical body weight estimation in small-scale livestock farms. Specifically, while the XGBoost algorithm demonstrated statistically significant and high performance on the training data, this performance was not sufficiently reflected in the test data, increasing the risk of overfitting in limited data structures. When multicollinearity, a potential factor that can lead to overfitting, was controlled, the XGBoost algorithm’s sensitivity to complex variable interactions may have led to performance degradation in limited sample sizes. In contrast, thanks to its more flexible structure and high-variance sampling mechanism, the Random Forest algorithm overcame these limitations and produced more consistent and generalizable results across both training and test sets. However, the study has several limitations. The limited sample size, the fact that data were obtained from only a specific geographic region, and the comparison of only two algorithms may limit the generalizability of the results to different populations. Furthermore, the limited range of hyperparameter optimization may have prevented the models from reaching their full potential performance. Despite these limitations, the study’s findings point to low-cost, applicable model structures that can be integrated into decision-support processes. Future research will focus on assessing different species, utilizing larger and more diverse datasets, and examining diverse environmental conditions. Furthermore, by integrating deep learning and image processing-based methods, we aim to increase model accuracy and generalizability, enabling the development of more effective prediction systems that contribute to animal welfare. Authors’ contributions Conceptualization, M.Á.G.D., C.T. and A.J.C.C.; methodology, M.Á.G.D., C.T., R.I.P.C., I.V.M., A.G.V., A.C.H., E.C.P., D.A.D.C., U.Ş., K.C., Y.A., H.T.Ç. and A.J.C.C.; software, M.Á.G.D., C.T., U.Ş. and A.J.C.C.; validation, M.Á.G.D., C.T., R.I.P.C., I.V.M., A.G.V., A.C.H., E.C.P., D.A.D.C., H.,Ö., U.Ş., K.C., Y.A., H.T.Ç., O.O. and A.J.C.C.; formal analysis, C.T., H.Ö., U.Ş. and A.J.C.C.; investigation, M.Á.G.D., C.T., R.I.P.C., I.V.M., A.G.V., A.C.H., E.C.P., D.A.D.C., U.Ş., K.C., Y.A., H.T.Ç., O.O. and A.J.C.C.; resources, M.Á.G.D., C.T., U.Ş. and A.J.C.C.; data curation, M.Á.G.D., C.T., R.I.P.C., I.V.M., A.G.V., A.C.H., E.C.P., D.A.D.C., U.Ş. and A.J.C.C.; writing—original draft preparation, M.Á.G.D., C.T., U.Ş. and A.J.C.C.; writing—review and editing, M.Á.G.D., C.T., R.I.P.C., I.V.M., A.G.V., A.C.H., E.C.P., D.A.D.C., U.Ş., K.C., Y.A., H.T.Ç. and A.J.C.C.; visualization, C.T., U.Ş. and A.J.C.C.; supervision, A.J.C.C.; project administration, M.Á.G.D., R.I.P.C., I.V.M., A.G.V., A.C.H., E.C.P., D.A.D.C. and A.J.C.C.; funding acquisition, M.Á.G.D. and A.J.C.C. All authors have read and agreed to the published version of the manuscript. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Data availability The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The experiment was approved by the Animal Ethics Committee of the Scientific Department of Agricultural Sciences, Autonomous University of Tabasco, in accordance with the Standards for Ethical Animal Research (record no. CIEI: Folio 1173–2022; dated September 29, 2022). The experiment was conducted on a privately owned farm. Prior to sampling, written informed consent was obtained from the farm owner, who provided signed authorization for the use of animals and farm facilities for research purposes. The study did not involve any invasive procedures, and all animal handling strictly complied with ethical standards and institutional guidelines for the care and use of animals in research. All experimental methods followed established ethical and regulatory protocols consistent with the ARRIVE guidelines. During the experimental period, all animals were clinically healthy. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes The original online version of this article was revised: Following publication of the original article [1], after carefully reviewing the final published version, the authors unfortunately noticed that several corrections clearly indicated in the proof file, which was revised together with the corresponding author, were not reflected in the published article. In addition, the order of the figures appears to have been mixed during production, and some figure captions do not correspond to the figures currently presented in the article. Since figures are essential for the accurate interpretation of the results, this issue may cause confusion for readers. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Change history 7/20/2026 The original online version of this article was revised: Following publication of the original article [1], after carefully reviewing the final published version, the authors unfortunately noticed that several corrections clearly indicated in the proof file, which was revised together with the corresponding author, were not reflected in the published article. In addition, the order of the figures appears to have been mixed during production, and some figure captions do not correspond to the figures currently presented in the article. Since figures are essential for the accurate interpretation of the results, this issue may cause confusion for readers. Change history 7/30/2026 A Correction to this paper has been published: 10.1186/s12917-026-05748-6 References 1. Celik S, Eyduran E, Karadas K, Tariq MM. Comparison of predictive performance of data mining algorithms in predicting body weight in Mengali Rams of Pakistan. Braz J Vet Res Anim Sci. 2017;46(11):863–72. 10.1590/S1806-92902017001100005. [ Google Scholar ] 2. Tirink C. Comparison of bayesian regularized neural network, random forest regression, support vector regression, and multivariate adaptive regression splines algorithms to predict body weight from biometrical measurements in thalli sheep. Kafkas Univ Vet Fak Derg. 2022;28(3):411–9. 10.9775/kvfd.2022.27164. [ Google Scholar ] 3. Estevez-Moreno LX, Sanchez-Vera E, Nava-Bernal G, Estrada-Flores JG, Gomez-Demetrio W, Sepúlveda WS. The role of sheep production in the livelihoods of Mexican smallholders: evidence from a park-adjacent community. Small Rumin Res. 2019;178:94–101. 10.1016/j.smallrumres.2019.08.001. [ Google Scholar ] 4. Magaña-Monforte JG, Huchin-Cab M, Ake-López JR, Segura-Correa JC. A field study of reproductive performance and productivity of pelibuey Ewes in southeastern Mexico. Trop Anim Health Prod. 2013;45:1771–6. 10.1007/s11250-013-0431-2. [ DOI ] [ PubMed ] [ Google Scholar ] 5. García-Cigarroa JC, Luna-Mendicuti AA, Canul-Solís JR, Castillo-Sanchez LE, Herrera-Camacho J, Vargas-Bello-Pérez E, et al. Use of real-time ultrasound measurements of fat thickness and longissimus thoracis muscle characteristics for predicting body fat depots in crossbred hair Ewes. Small Rumin Res. 2024;241:107400. 10.1016/j.smallrumres.2024.107400. [ Google Scholar ] 6. Salazar-Cuytun R, García-Herrera RA, Muñoz-Benitez AL, Camacho-Pérez E, Muñoz-Osorio GA, Ptacek M, et al. Relationship between body volume and body weight in pelibuey Ewes. Trop Subtrop Agroecosyst. 2021;24(3):1–6. [ Google Scholar ] 7. Almeida A. Barbados blackbelly: the Caribbean ovine genetic resource. Trop Anim Health Prod. 2017;2:239–50. 10.1007/s11250-017-1475-5. [ Google Scholar ] 8. Escalante-Clemente S, Vázquez-Jiménez S, López-Durán SK, Arcos-Alvarez DN, Arbez-Abnal TA, Piñeiro-Vazquez ÁT, et al. Using the 9th–11th rib section to predict carcass tissue composition in blackbelly sheep. Ital J Anim Sci. 2022;21(1):161–7. 10.1080/1828051X.2021.2002731. [ Google Scholar ] 9. Vazquez-Martinez I, Tirink C, Salazar-Cuytun R, Mezo-Solis JA, Garcia-Herrera RA, Orzuna-Orzuna JF, et al. Predicting body weight through biometric measurements in growing hair sheep using data mining and machine learning algorithms. Trop Anim Health Prod. 2023;55(5):307–16. 10.1007/s11250-023-03717-x. [ DOI ] [ PubMed ] [ Google Scholar ] 10. Chay-Canul AJ, Tapia-González J, Canul-Solís JR, Casanova-Lugo F, Piñeiro-Vázquez ÁT, Portillo-Salgado R, et al. Predictive biometrics of hair sheep through digital imaging. Vet Mex. 2023;10(1):11. 10.22201/fmvz.24486760e.2023.1150. [ Google Scholar ] 11. Chay-Canul AJ, Camacho-Perez E, Casanova-Lugo F, Rodriguez-Abreo O, Cruz-Fernandez M, Rodriguez-Resendiz J. Neural Network-Based body weight prediction in pelibuey sheep through biometric measurements. Technol. 2024;12:59. 10.3390/technologies12050059. [ Google Scholar ] 12. Wishart H, Morgan-Davies C, Stott AW, Wilson A, Waterhouse T. Liveweight loss associated with handling and weighing of grazing sheep. Small Rumin Res. 2017;153:163–70. 10.1016/j.smallrumres.2017.06.013. [ Google Scholar ] 13. Fuentes S, Viejo CG, Tongson E, Dunshea FR. The livestock farming digital transformation: implementation of new and emerging technologies using artificial intelligence. Anim Health Res Rev. 2022;23(1):59–71. 10.1017/S1466252321000177. [ DOI ] [ PubMed ] [ Google Scholar ] 14. Khan MA, Tariq MM, Eyduran E, Tatliyer A, Rafeeq M, Abbas F, et al. Estimating body weight from several body measurements in Harnai sheep without multicollinearity problem. J Anim Plant Sci. 2014;24(1):120–6. [ Google Scholar ] 15. Ali M, Eyduran E, Tariq MM, Tirink C, Abbas F, Bajwa MA, et al. Comparison of artificial neural network and decision tree algorithms used for predicting live weight at post-weaning period from some biometrical characteristics in Harnai sheep. Pak J Zool. 2015;47:1579–85. [ Google Scholar ] 16. Olfaz M, Tirink C, Önder H. Use of CART and CHAID algorithms in Karayaka sheep breeding. Kafkas Univ Vet Fak Derg. 2019;25(1):105–10. 10.9775/kvfd.2018.20388. [ Google Scholar ] 17. Sabbioni A, Beretti V, Superchi P, Ablondi M. Body weight Estimation from body measures in cornigliese sheep breed. Ital J Anim Sci. 2020;19:25–30. 10.1080/1828051X.2019.1689189. [ Google Scholar ] 18. Ruchay A, Kober V, Dorofeev K, Kolpakov V, Dzhulamanov K, Kalschikov V, et al. Comparative analysis of machine learning algorithms for predicting live weight of Hereford cows. Comput Electron Agric. 2022;195:106837. 10.1016/j.compag.2022.106837. [ Google Scholar ] 19. Salazar-Cuytun R, Portillo-Salgado R, García-Herrera RA, Camacho-Pérez E, Zaragoza-Vera CV, Gurgel ALC, et al. Prediction of live weight in growing hair sheep using the body volume formula. Arq Bras Med Vet Zootec. 2022;74:483–9. 10.1590/1678-4162-12624. [ Google Scholar ] 20. Chen T, Guestrin C. Xgboost: A scalable tree boosting system. Proc 22nd ACM SIGKDD Int Conf Knowl Discov Data Min. 2016;13–17. 10.1145/2939672.293978. 21. Punuri SB, Kuanar SK, Kolhar M, Mishra TK, Alameen A, Mohapatra H, et al. Efficient Net-XGBoost: an implementation for facial emotion recognition using transfer learning. Math. 2023;11(3):776. 10.3390/math11030776. [ Google Scholar ] 22. Faraz A, Tirink C, Önder H, Şen U, Ishaq HM, Tauqir NA, et al. Usage of the XGBoost and MARS algorithms for predicting body weight in Kajli sheep breed. Trop Anim Health Prod. 2023;55(4):276. 10.1007/s11250-023-03700-6. [ DOI ] [ PubMed ] [ Google Scholar ] 23. Sagi O, Rokach L. Approximating XGBoost with an interpretable decision tree. Inf Sci. 2021;572:522–42. 10.1016/j.ins.2021.05.055. [ Google Scholar ] 24. Gertz M, Grobe-Butenuth K, Junge W, Maassen-Francke B, Renner C, Sparenberg H, et al. Using the XGBoost algorithm to classify neck and leg activity sensor data using on-farm health recordings for locomotor-associated diseases. Comput Electron Agric. 2020;173:105404. 10.1016/j.compag.2020.105404. [ Google Scholar ] 25. Breiman L. Random forests. Mach Learn. 2001;45(1):5–32. 10.1023/a:1010933404324. [ Google Scholar ] 26. Rodriguez-Galiano V, Mendes MP, Garcia-Soldado MJ, Chica-Olmo M, Riberio L. Predictive modeling of groundwater nitrate pollution using random forest and multisource variables related to intrinsic and specific vulnerability: A case study in an agricultural setting (Southern Spain). Sci Total Environ. 2014;476:189–206. 10.1016/j.scitotenv.2014.01.001. [ DOI ] [ PubMed ] [ Google Scholar ] 27. Wang L, Zhou X, Zhu X, Dong Z, Guo W. Estimation of biomass in wheat using random forest regression algorithm and remote sensing data. Crop J. 2016;4:212–9. 10.1016/j.cj.2016.01.008. [ Google Scholar ] 28. Tirink C, Piwczyński D, Kolenda M, Önder H. Estimation of body weight based on biometric measurements by using random forest regression, support vector regression and CART algorithms. Animals. 2023;13(5):798. 10.3390/ani13050798. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Liaw A, Wiener M. Classification and regression by randomforest. R News. 2002;2:18–22. [ Google Scholar ] 30. R Core Team. R: A Language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2022. [ Google Scholar ] 31. Van Rossum G, Drake FL. The python Language reference. Amsterdam: Python Software Foundation; 2010. [ Google Scholar ] 32. Revelle W. Psych: procedures for personality and psychological research. Evanston: Northwestern University; 2015. [ Google Scholar ] 33. Wei T, Simko V, Levy M, Xie Y, Jin Y, Zemla J. Package corrplot. Statistician. 2017;56:e24. [ Google Scholar ] 34. Kuhn M. Caret: classification and regression training. R Package Version. 2022;6:0–93. [ Google Scholar ] 35. Chen T, He T, Benesty M, Khotilovich V, Tang Y, Cho H et al. xgboost: Extreme gradient boosting. R package version 1.7.3. 2023;1. 36. Zhou ZH. Ensemble methods: foundations and algorithms. CRC; 2025. 37. Canaza-Cayo AW, Churata-Huacani R, Çakmakçı C, Rodríguez-Huanca FH, de Sousa Bueno Filho JS, Fernandes TJ, De La Cruz YCR. Use of machine learning approaches for body weight prediction in Peruvian Corriedale sheep. Smart Agricultural Technol. 2024;7:100419. [ Google Scholar ] 38. Alsahaf A, Azzopardi G, Ducro B, Veerkamp RF, Petkov N. Predicting slaughter weight in pigs with regression tree ensembles. In: Petkov N, Strisciuglio N, Travieso-González CM, editors. Applications of intelligent systems: Proceedings of the 1st International APPIS Conference. Amsterdam: IOS Press BV; 2018. pp. 1–9. 10.3233/978-1-61499-929-4-1 39. Hamadani A, Ganai NA. Artificial intelligence algorithm comparison and ranking for weight prediction in sheep. Sci Rep. 2023;13(1):13242. 10.1038/s41598-023-40528-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Kozaklı Ö, Ceyhan A, Noyan M. Comparison of machine learning algorithms and multiple linear regression for live weight Estimation of Akkaraman lambs. Trop Anim Health Prod. 2024;56(7):250. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Megersa AG, Negash F, Hailu A, Melak A, Assefa A, Aseged T, Sinkie S. (2025). Prediction of body weight in Indigenous sheep using random forest regression, support vector regression, and classification and regression trees algorithms. Veterinary Anim Sci, 100517. 42. Kunene NW, Nesamvuni AE, Nsahlai IV. Determination of prediction equations for estimating body weight of Zulu (Nguni) sheep. Small Ruminant Res. 2009;84(1–3):41–6. [ Google Scholar ] 43. Castillo PE, Macedo RJ, Arredondo V, Zepeda JL, Valencia-Posadas M, Haubi CU. Morphological description and live weight prediction from body measurements of Socorro Island Merino Lambs. Animals. 2023;13(12):1978. 44. Çam M, Olfaz M, Soydan E. Body measurements reflect body weights and carcass yields in Karayaka sheep. Asian J Anim Veterinary Adv. 2010;5(2):120–127. 45. Vázquez-Martínez I, Tırınk C, Salazar-Cuytun R, Mezo-Solis JA, Garcia Herrera RA, Orzuna-Orzuna JF, Chay-Canul AJ. Predicting body weight through biometric measurements in growing hair sheep using data mining and machine learning algorithms. Trop Anim Health Prod. 2023;55(5):307. [ DOI ] [ PubMed ] [ Google Scholar ] 46. Posbergh CJ, Huson HJ. All sheeps and sizes: a genetic investigation of mature body size across sheep breeds reveals a polygenic nature. Anim Genet. 2021;52(1):99–107. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Coşkun G, Şahin Ö, Altay Y, Aytekin İ. Final fattening live weight prediction in Anatolian Merinos lambs from some body characteristics at the initial of fattening by using some data mining algorithms. Black Sea J Agric. 2023;6(1):47–53. 10.47115/bsagriculture.1181444. [ Google Scholar ] 48. Esener N, Eşki HT. Early prediction of final weight in nomadic sheep flocks using machine learning algorithms. Trop Anim Health Prod. 2025;57(8):420. [ DOI ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Articles from BMC Veterinary Research are provided here courtesy of BMC ACTIONS View on publisher site PDF (1.6 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top