A comparative evaluation of machine learning algorithms for predicting chicken egg weight from egg quality traits - PMC 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 Vet Anim Sci . 2026 Mar 30;32:100641. doi: 10.1016/j.vas.2026.100641 Search in PMC Search in PubMed View in NLM Catalog Add to search A comparative evaluation of machine learning algorithms for predicting chicken egg weight from egg quality traits Ashenafi Getachew Megersa Ashenafi Getachew Megersa 1 School of Animal and Range Sciences, Haramaya University, P. O. Box 138, Dire Dawa, Ethiopia Find articles by Ashenafi Getachew Megersa 1 Author information Article notes Copyright and License information 1 School of Animal and Range Sciences, Haramaya University, P. O. Box 138, Dire Dawa, Ethiopia Collection date 2026 Jun. © 2026 The Author This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). PMC Copyright notice PMCID: PMC13090620 PMID: 42007391 Abstract This study compared the effectiveness of artificial neural network (ANN), K-Nearest Neighbors (KNN), Gradient Boost regression (GBoost), random forest regression (RFR), support vector regression (SVR), and classification and regression trees (CART) algorithms for developing predictive models to estimate the egg weight of Bovan brown chickens in Ethiopia based on egg quality traits. Data were collected from 600 consumption eggs at the Haramaya University poultry farm, with 300 eggs from each of the cage and deep litter production systems. The dataset included egg weight (EW) and ten egg quality metrics. Descriptive statistics revealed the strongest correlations with egg weight were albumen weight (AW) and egg length (EL), with correlation coefficients (r) of 0.86 and 0.64, respectively. Performance among the models varied, with GBoost achieving the highest predictive accuracy. It yielded the highest coefficient of determination (R² = 0.960), the lowest root means square error (1.112), Akaike information criterion (AIC = 1834.394), and Bayesian information criterion (BIC = 1847.544), as well as the highest correlation (0.98) between observed and predicted values. In contrast, the CART model demonstrated the weakest overall performance. A relative importance analysis identified AW as the most significant predictor across all models, accounting for 96.28 % in CART, 80.56 % in GBoost, 72.69 % in RFR, 30.6 % in ANN, 29.9 % in SVR, and 29.1 % in KNN. The results indicate that the GBoost algorithm is a reliable and superior method for predicting egg weight. This study suggests that, with the application of this model, egg producers can effectively anticipate egg weight based on key egg quality characteristics. Keywords: Egg weight prediction, Egg quality traits, Relative importance, Machine learning algorithms, Predictive performance Introduction The chicken egg is a remarkable source of nutrition, providing a well-balanced profile of essential nutrients and high-quality protein, while being relatively low in energy content ( Akyurek & Okur, 2009 ; Rdhault-Godbert et al., 2019 ; Stadelman & Cotterill, 2017 ). As a widely consumed animal product, eggs are considered a complete food for a majority of the global population ( Baishya et al., 2008 ). For consumers, specific egg characteristics are critical for product acceptance, and for producers, these traits are vital indicators of commercial quality, influencing grading, hatchability, and subsequent chick performance ( Stadelman & Cotterill, 2017 ). Key external and internal egg parameters—such as egg weight, albumen weight, yolk weight, shell thickness, and various dimensional measurements—directly impact a business's outcomes, including hatchling weight and viability ( Farooq et al., 2001 ; Wilson & Suarez, 1993 ). An accurate egg weight prediction is therefore essential. For breeding programs, predicting egg weight prior to incubation allows for the anticipation of chick weight, a critical factor for genetic improvement strategies ( Wilson & Suarez, 1993 ). This is particularly important given the well-established direct relationship between egg weight and the weight of the newly hatched chick ( Farooq et al., 2001 ). Consequently, policymakers and breeders require reliable methods to estimate egg weight to make informed decisions. For instance Wu, X. et al. (2022) demonstrate that adaptive federated learning (FL) models surpass conventional centralized or local learning approaches under fixed communication budgets. This finding has significant implications for industries like poultry production, where privacy-preserving FL can enable secure cross-farm collaboration. By training a shared global model on decentralized data, this approach directly mitigates two critical issues: model overfitting (to a single farm's environment) and model volatility (unpredictable performance across heterogeneous farm conditions). However, in Ethiopia, research on egg quality has been limited ( Biazin, 2021 ; Mummed et al., 2024 ; Wondmeneh, 2015 ), with studies often relying on small sample sizes and univariate statistical analyses, such as analysis of variance and simple correlations. A significant limitation of univariate approaches is that they examine variables in isolation, failing to account for the complex biological interdependencies between egg quality traits, which often arise from pleiotropy or genetic linkage. This can lead to redundant findings and an incomplete understanding of the system ( Yang et al., 2006 ). In contrast, data mining techniques offer a powerful alternative for analyzing complex agricultural datasets. These methods have been successfully applied in animal science to predict quail egg weights from internal and external characteristics ( Çelik et al., 2017 ), investigate factors affecting fertility in Japanese quail ( Çelik et al., 2016 ), and estimate body and carcass weight in livestock using biometric data ( Ali et al., 2015 ; Karabacak et al., 2017 ; Tyasi et al., 2020 ). Algorithms like Classification and Regression Trees (CART) and Chi-squared Automatic Interaction Detection (CHAID) have also been effectively used in sheep breeding programs ( Olfaz et al., 2019 ). Despite this demonstrated utility in related domains, the application of advanced data mining for the analysis of chicken egg traits for breeding and selection remains largely unexplored in Ethiopia. To address this research gap, the present study was conducted with the primary objective of evaluating and comparing the performance of multiple data mining algorithms—including Artificial Neural Network (ANN), K-Nearest Neighbors (KNN), Random Forest Regression (RFR), Support Vector Regression (SVR), Gradient Boost (GBoost), and CART—in predicting the egg weight of Bovan brown chickens from multiple measurements of egg quality traits, including yolk weight (YW), shell weight (SW), albumen weight (AW), albumen height (AH), yolk height (YH), shell thickness (ST), egg length (EL), egg width (EWi), and yolk color (YC). By identifying the most accurate predictive model, this research aims to provide a reliable tool for egg weight estimation of consumption eggs from egg quality traits, thereby supporting more effective breeding and selection programs in the Ethiopian poultry industry. Materials and methods Study area The research was carried out at the Haramaya University Poultry Farm, located 505 km east of Addis Ababa. The geographic coordinates of the site are 9°26′ N latitude and 42°03′ E longitude, at an elevation of 1980 m above sea level. The region receives an average annual rainfall of 741.6 mm. The average annual maximum and minimum temperatures are 23.4 °C and 8.25 °C, respectively. Experimental animal and data collection This study utilized Bovan brown chickens. All birds were managed under uniform conditions and fed a standard diet formulated with the following ingredients: maize (64 %), noug (Guizotiaabyssinica) seed cake (13 %), wheat middling (3 %), soybean meal (16.19 %), dicalcium phosphate (1.2 %), limestone (1.4 %), salt (0.5 %), DL-methionine (0.05 %), L-lysine HCl (0.65 %), and a general vitamin-mineral premix (0.01 %). The birds were provided with feed and clean drinking water ad libitum, with the dietary regimen meeting established nutritional standards: a starter diet containing 20 % crude protein (CP) and 2800 kcal/kg metabolizable energy (ME) for the first eight weeks, a grower diet with 16 % CP and 2800 kcal/kg ME from 9 to 20 weeks, and a layer diet with 16.5 % CP and 2750 kcal/kg ME during the laying phase. A standard vaccination protocol was followed, protecting against Marek's disease (day 1), Newcastle disease (days 3, 28, 63, and 112), Gumboro disease (days 14 and 21), fowl pox (day 72), and fowl typhoid (days 45 and 84). The external and the internal egg quality traits were measured. External egg qualities were measured before breaking open the eggs. The egg weight (g) was determined on a digital weighing scale, and the length and width of an egg were measured with Vernier calipers. The eggs were cracked open on the egg-breaking stand to measure their internal traits after their external qualities had been measured. The internal traits of eggs, such as shell thickness (mm), after the eggshell was rinsed to remove any albumin that had adhered to it, were determined using a shell thickness gauge. The shell weight (g) was obtained after removing the egg's internal content. The yolk weight was measured after adhering albumin was removed by rolling the yolks over filter paper, and then. The difference between egg weight and shell weight and yolk weight was used to calculate the albumin weight (g). Albumin’s height was measured using a spherometer on a glass surface. A spirometer was used on a glass surface to measure yolk height (mm). The DSM yolk color fan was used to provide a score on a range of 1 to 16. A total of 600 eggs were collected for this study, with 300 eggs sourced from a deep litter production system and 300 from a cage system. Eggs were gathered from hens aged 42 to 72 weeks. All egg quality measurements were performed on the day of collection. The following traits were recorded: egg weight (EW), yolk weight (YW), albumen weight (AW), shell weight (SW), albumen height (AH), Yolk height (YH), egg length (EL), egg width (EWi), shell thickness (ST), and yolk colour (YC). Statistical analysis Descriptive statistics, including mean, standard deviation, standard error, coefficient of variation, minimum, and maximum, were calculated for all egg quality traits (EW, AH, AW, YH, YW, SW, YC, EL, EWi, and ST) using JMP Pro version 18 ( SAS Institute, 2023 ). Six machine learning algorithms were employed to model the relationship between these traits and egg weight: ANN, KNN, RFR, SVR, CART, and GBoost. Brief descriptions of these algorithms are provided below. To ensure a comprehensive and fair evaluation a collected data set was implemented for this research because, the cross-layer metric framework proposed by He (2025) —encompassing model performance, system efficiency, and operational cost—was adopted. Consequently, all comparative models, including the adaptive federated learning approach and conventional baselines, were trained and evaluated on the same collected dataset using this unified framework. Gradient Boosting (GBoost): Gradient Boosting was employed as a second ensemble method for predicting egg weight (EW). In contrast to the parallel, bagging-based approach of Random Forest, Gradient Boosting is a sequential, additive technique that builds a strong predictive model by iteratively combining multiple weak learners; typically shallow decision trees ( Friedman, 2001 ). The algorithm begins with a simple initial model (e.g., the mean of the target variable). In each subsequent iteration, a new weak learner is trained not on the original target, but on the residual errors—the differences between the current ensemble's predictions and the true values. This new tree is then added to the ensemble, with its contribution scaled by a learning rate (a shrinkage parameter). This process focuses each new model on correcting the mistakes of the combined predecessors, gradually minimizing the overall prediction error in a stage-wise manner ( Freeman et al., 2016 ; Jun, 2021 ). The learning rate imposes partial shrinkage on the contribution of each tree, which acts as a form of regularization to prevent overfitting and often leads to superior accuracy. Furthermore, as the model integrates explanatory variables sequentially to explain residual variance, the frequency and contribution of each variable can be analyzed to provide a robust measure of relative feature importance for predicting EW, offering insights distinct from permutation-based methods ( González-Recio et al., 2013 ). Artificial Neural Network (ANN): Artificial Neural Networks (ANNs) were implemented as a powerful, nonlinear modeling technique to predict egg weight (EW). ANNs are particularly advantageous for addressing complex problems where the underlying relationships between input and output variables are unknown or difficult to specify with traditional parametric models. A key strength is their ability to learn these relationships directly from data without a predefined functional form, acting as universal function approximators ( Mittal & Zhan, 2000 ). This capability has led to their successful application in various agricultural modeling tasks, including the prediction of nutritional components and animal performance ( Cravener & Roush, 1999 ; Edriss et al., 2008 ; Roush et al., 1997 ). Furthermore, ANNs can be designed to predict multiple dependent variables simultaneously, offering flexibility beyond single-output models . In this study, a supervised, feed forward multilayer perceptron (MLP) architecture was employed. The fundamental unit is an artificial neuron (perceptron), which receives weighted inputs, sums them, and passes the result through a non-linear activation function to produce an output. Networks are structured into layers: an input layer (egg traits), one or more hidden layers, and an output layer (EW). The model was trained using a supervised learning paradigm. The network was iteratively presented with known input-output pairs. Through an optimization algorithm, typically back propagation with gradient descent, the internal connection weights between neurons were adjusted to minimize a loss function (e.g., mean squared error) between the network's predictions and the actual target values. This iterative cycle over the dataset, known as an epoch, was repeated until the model's error converged to a minimum, indicating the network was trained. Once training was complete, the final optimized set of weights was fixed. This trained network could then process new, unseen input data to generate predictions for EW, effectively generalizing the complex, non-linear patterns learned from the training data. K-Nearest Neighbor (KNN): The K-Nearest Neighbors (KNN) algorithm is a supervised, instance-based learning method ( Cover & Hart, 1967 ). It operates on the principle of similarity, making it a non-parametric technique suitable for both classification and regression. In this study, it was applied as a regressor to predict egg weight (EW). For a new data point (defined by its egg traits), the algorithm identifies the *k* most similar instances—the "nearest neighbors"—from the training set. Similarity is quantified by a distance metric, with Euclidean distance being the standard measure ( Danil et al., 2019 ; Sökün et al., 2012 ). The predicted EW for the new instance is calculated as the mean of the EW values from these *k* neighbors. The integer *k* is the algorithm's critical hyperparameter, controlling model complexity by determining the size of the local neighborhood used for prediction. A small *k* (e.g., 1) results in a flexible model sensitive to local noise, while a larger *k* yields smoother, more stable predictions by averaging across a broader region. The optimal value of *k* for this regression task was determined through cross-validation to minimize prediction error. Random Forest (RFR): The Random Forest algorithm, an ensemble learning method, was employed to predict egg weight (EW). This technique enhances predictive accuracy and robustness by constructing a large collection of de-correlated decision trees ( Breiman, 2001 ). Each tree is trained on a random bootstrap sample of the observations (bagging), and at each node split, the algorithm evaluates only a random subset of the predictor variables. The final prediction is generated by aggregating—averaging, in this regression context—the predictions from all individual trees. This ensemble approach effectively reduces model variance and mitigates overfitting, making it suitable for capturing complex relationships within the data ( Huang et al., 2020 ; Reis et al., 2018 ; Shahinfar et al., 2019 ). Key hyperparameters were optimized to ensure model performance. The ensemble size was fixed at 200 trees based on preliminary analysis indicating stabilized error estimates. A grid-search procedure with cross-validation was used to optimize the complexity of the individual trees, specifically the maximum depth and the minimum number of samples required to split a node. In line with the heuristic suggested by Breiman (2001) , the number of features considered at each split was set to the square root of the total number of predictors (i.e., 4). All available features were used for model training to allow for a subsequent assessment of their relative importance. The contribution of each egg trait to the model's predictive accuracy was quantified using permutation importance ( Altmann et al., 2010 ). This method evaluates the mean decrease in the model's accuracy when the values of a specific feature are randomly shuffled, thereby breaking its relationship with the target variable. A greater drop in accuracy indicates that the feature is more critical for the model's performance. This process was repeated for each predictor to rank all features by their estimated importance for predicting EW. Support Vector Regression (SVR): Support Vector Regression (SVR) was implemented as a robust machine-learning algorithm for predicting egg weight (EW). SVR is well-regarded for its capacity to model complex, nonlinear relationships, perform effectively with smaller datasets, and reliably converge to a global optimum by formulating the learning task as a convex optimization problem ( Brasil et al., 2022 ; Chen et al., 2023 ; Gammermann, 2000 ; Kecman, 2001 ). The algorithm seeks to identify a predictive function, *f*( x ), that best approximates the continuous target variable. Its core innovation is the use of an ε-insensitive tube (or margin) around the regression function. The optimization procedure aims to fit the flattest possible function while ensuring that the deviation of predicted values from the actual observations yᵢ does not exceed ε for the majority of training data points ( Peng & Xu, 2016 ; Canaza-Cayo et al., 2024 ). This dual objective balances model simplicity (flatness) for improved generalization with predictive accuracy, tolerating minor errors within the specified margin. To capture intricate, non-linear patterns between egg traits and EW, SVR utilizes kernel functions. These functions implicitly project the input data into a higher-dimensional feature space where a linear regression hyperplane can be constructed to model the relationships. This "kernel trick" allows SVR to function as a powerful nonparametric regression tool without the computational burden of explicit high-dimensional transformations, effectively modeling complex feature interactions. Classification and Regression Tree (CART): To predict egg weight (EW) from physical egg traits, the Classification and Regression Tree (CART) algorithm was employed ( Breiman et al., 1984 ). This non-parametric, recursive partitioning method constructs a decision tree model by repeatedly splitting the dataset into increasingly homogeneous subgroups with respect to the target variable, EW. Starting from a root node containing all observations, the algorithm repeatedly identifies optimal cut-off values for the traits to create binary splits. Each split aims to produce subsets (nodes) that are more homogeneous in EW than their parent node. This process continues until terminal nodes are formed, yielding an interpretable tree structure that captures both linear and non-linear relationships. To ensure the model generalizes well, overfitting was mitigated through rigorous model selection. The optimal tree size was determined using 10-fold cross-validation combined with the one-standard-error rule. The final model was selected based on a combination of performance metrics: Akaike’s information criterion (AICc), Bayesian information criteria (BIC), Root means square error (RMSE), Mean absolute error (MAE), Mean squared error (MSE), Coefficient of determination (R2), Adjusted R-squared (AdjR2), Pearson’s correlation coefficients (r). Additionally, statistical significance of splits was adjusted using the Bonferroni method to account for multiple comparisons. Model performance and complexity were evaluated using the corrected Akaike Information Criterion (AICc). The final model selection was based on several metrics: Bayesian Information Criterion (BIC), Root Mean Square Error (RMSE), R-squared (R²), Adjusted R-squared (AdjR²), correlation coefficient (r), and AICc. Significance thresholds for splitting were adjusted using the Bonferroni method ( Breiman et al., 1984 ; Mikail & Bakır, 2019 ; Ouadah et al., 2022 ). Results Descriptive statistics and correlation analysis Descriptive statistics and correlation analyses were performed to summarize the egg quality traits and to examine their linear relationships with EW. The strength of the linear relationships was assessed using Pearson correlation coefficients. Descriptive statistics were computed from a sample of 600 eggs, comprising 300 eggs from a deep litter system and 300 from a cage system ( Table 1 ). The mean EW was 63.46 g. Among the other parameters, AW was the highest (mean = 42.04 g), while YC had the lowest mean value (1.23). The coefficient of variation (CV) indicated that YC was highly variable (CV = 51.89 %), while SW and ST showed moderate variability (CV between 15 % and 30 %). The remaining traits—egg weight (EW), albumen height (AH), albumen weight (AW), yolk height (YH), yolk weight (YW), egg length (EL), and egg width (EWi)—exhibited low variability (CV < 15 %). Table 1. Descriptive statistics for egg quality traits ( n = 600). Traits Mean CV SD SE Minimum Maximum EW(g) 63.46 8.75 5.55 0.23 48.96 89.4 AH (mm) 10.85 16.23 1.76 0.07 5 15.6 AW(g) 42.04 10.63 4.47 0.18 29.62 59.3 YH(mm) 14.36 6.74 0.97 0.04 10 19 YW(g) 15.71 9.42 1.48 0.06 9.12 19.4 SW(g) 4.99 20.06 1.00 0.04 2.84 7.9 EL(mm) 5.45 4.96 0.27 0.01 4.7 6.5 EWi(mm) 4.23 5.68 0.24 0.01 2.9 5.4 ST(mm) 0.81 26.93 0.22 0.01 0.17 1.78 YC(-) 1.23 51.89 0.64 0.03 1 6 Open in a new tab egg weight (EW), albumen height (AH), albumen weight (AW), yolk height (YH), yolk weight (YW), shell weight (SW), egg length (EL), egg width (EWi), shell thickness (ST), yolk colour (YC), coefficient of variation (CV), standard deviation (SD), standard error (SE). Correlation analysis revealed significant relationships between EW and all egg quality parameters except for YC ( Table 2 ). AW demonstrated the strongest positive correlation with EW ( r = 0.86) and this indicates egg weight and albumen weight clearly have a very significant link, which means that finding larger eggs would be directly related to albumen weight. EL, EWi, and SW also showed strong positive correlations ( r = 0.64, 0.53, and 0.47, respectively). The remaining parameters exhibited moderate correlations with EW. Table 2. Correlation coefficients between egg weight and egg quality parameters ( n = 600). Traits EW AH AW YH YW SW EL EW ST YC EW 1.00 AH 0.10* 1.00 AW 0.86* 0.19* 1.00 YH 0.22* 0.01* 0.16* 1.00 YW 0.44* −0.09 ns 0.21* 0.28* 1.00 SW 0.47* −0.04* 0.30* 0.19* 0.21* 1.00 EL 0.64* 0.06 ns 0.60* 0.19* 0.30* 0.38* 1.00 EWi 0.53* 0.25 ns 0.52* 0.23* 0.17* 0.44* 0.54* 1.00 ST 0.20* −0.04 ns 0.08* 0.13* 0.12* 0.66* 0.20* 0.30* 1.00 YC 0.06 ns −0.24* −0.00 ns 0.09* 0.07 ns 0.04 ns −0.02 ns −0.14* −0.06 ns 1.00 Open in a new tab egg weight (EW), albumen height (AH), albumen weight (AW), yolk height (YH), yolk weight (YW), shell weight (SW), egg length (EL), egg width (EWi), shell thickness (ST), yolk color (YC). Predictive performance of the algorithms The CART algorithm produced a regression tree with 14 nodes using three primary predictors: AW, YW, and SW ( Fig. 1 ). The root node comprises the whole dataset, with an average egg weight of 63.46 g and a standard deviation (SD) of 5.55. The primary divisions were based on albumen weight (AW), leading to significant branches at different levels of AW. Fig. 1. Open in a new tab Regression tree of CART algorithm: AW, albumen weight; YW, yolk weight; SW, shell weight. For eggs with an albumen weight <38.61 g, the average egg weight was 57.09 g (SD = 3.47), covering 122 observations. When AW was 38.61 g or more, the average egg weight was 62.0 g (SD = 3.05, Logworth = 11.02) from 241 observations. This node was further split at a yolk weight (YW) of 15.34 g. For YW below 15.34 g, the average egg weight was 60.56 g (SD = 3.66) with 100 observations. For YW of 15.34 g or more, the mean weight increased to 63.02 g (SD = 1.98, Logworth=9.55) with 141 observations, and was further divided at AW = 41.33 g. For AW below 41.33 g, the mean weight was 62.34 g (SD = 1.84) from 92 observations, while for AW of 41.33 g or more, the mean egg weight increased to 64.29 g (SD=1.59). Another major branch (Node 2) split at AW = 49.2 g. Eggs with AW under 49.2 g had a mean weight of 67.0 g (SD = 2.77, Logworth= 28.37) with 202 observations. For AW of 49.2 g or more, the mean weight rose to 75.32 g (SD = 3.04) with 35 observations. Furthermore, Node 5 divided at AW=44.98 g. Eggs with AW <44.98 g had a mean weight of 65.37 g (SD = 2.11) with 101 observations. For AW of 44.98 g or more, the mean weight increased to 68.62 g (SD = 2.38, Logworth = 10.45) with 101 observations, and was finally split at a shell weight (SW) of 6.02 g. Eggs with SW below 6.02 g had a mean weight of 67.91 g (SD = 2.08) with 78 observations, while SW of 6.02 g or more resulted in a mean weight of 71.01 g (SD = 1.68) with 23 observations. A comparison of model performance indicated that GBoost was the most effective algorithm ( Table 3 ). It achieved the highest coefficient of determination (R² = 0.960) and correlation between observed and predicted values ( r = 0.98), along with the lowest error metrics (RMSE = 1.112, AIC = 1834.394, BIC = 1847.544). RFR also demonstrated strong, competitive performance (R² = 0.93, r = 0.965, RMSE = 1.467). SVR and ANN showed moderate performance, while KNN was less effective. The CART model demonstrated the weakest performance across all evaluated metrics (R² = 0.764, r = 0.874, RMSE = 2.701). Table 3. Predictive performance of CART, ANN, KNN, RFR, SVR and GBoost. Criteria CART ANN KNN RFR SVR GBoost Decision AICc 2898.994 2582.375 2709.865 2166.819 2373.926 1834.394 Smaller is better BIC 2912.144 2595.526 2723.016 2179.969 2387.077 1847.544 Smaller is better RMSE 2.701 2.074 2.307 1.467 1.744 1.112 Smaller is better MAE 1.839 1.157 1.701 0.747 0.803 0.635 Smaller is better MSE 2.696 2.080 2.519 1.491 1.753 1.120 Smaller is better R2 0.764 0.861 0.828 0.930 0.901 0.960 Greater is better AdjR2 0.763 0.860 0.827 0.930 0.901 0.960 Greater is better R 0.874 0.928 0.910 0.965 0.950 0.980 Greater is better Open in a new tab Akaike’s information criterion (AICc), Bayesian information criteria (BIC), Root means square error (RMSE), Mean absolute error (MAE), Mean squared error (MSE), Coefficient of determination (R2), Adjusted R-squared (AdjR2), Pearson’s correlation coefficients (r). Variable importance analysis identified AW as the most significant predictor across all models, though its relative contribution varied. Its importance was highest in the CART model (96.28 %; Fig. 2 ), followed by GBoost (80.56 %; Fig. 7 ) and RFR (72.69 %; Fig. 5 ). In the ANN ( Fig. 3 ), SVR ( Fig. 6 ), and KNN ( Fig. 4 ) models, AW remained the top predictor but with lower contributions (30.6 %, 29.9 %, and 29.1 %, respectively). Fig. 2. Open in a new tab Relative importance for CART algorithm: albumen weight (AW), yolk weight (YW), shell weight (SW), albumen height (AH), yolk height (YH), egg length (EL), egg width (EWi), shell thickness (ST), yolk color (YC). Fig. 7. Open in a new tab Relative importance for GBoost algorithm: albumen weight (AW), yolk weight (YW), shell weight (SW), egg length (EL), shell thickness (ST), egg width (EWi), albumen height (AH), yolk height (YH), yolk color (YC). Fig. 5. Open in a new tab Relative importance for RFR algorithm: albumen weight (AW), yolk weight (YW), egg length (EL), egg width (EWi), shell weight (SW), shell thickness (ST), yolk height (YH), albumen height (AH), yolk color (YC). Fig. 3. Open in a new tab Relative importance for ANN algorithm: albumen weight (AW), egg length (EL), yolk color (YC), egg width (EWi), shell weight (SW), yolk weight (YW), yolk height (YH), shell thickness (ST), albumen height (AH). Fig. 6. Open in a new tab Relative importance for SVR algorithm: albumen weight (AW), egg length (EL), yolk color (YC), egg width (EWi), yolk weight (YW), shell weight (SW), shell thickness (ST), yolk height (YH), albumen height (AH). Fig. 4. Open in a new tab Relative importance for KNN algorithm: albumen weight (AW), egg length (EL), egg width (EWi), yolk weight (YW), shell weight (SW), yolk color (YC), yolk height (YH), albumen height (AH), shell thickness (ST). YW was the second most important variable in the GBoost, RFR, and CART models. In contrast, EL was the second most important variable in the ANN, SVR, and KNN models. YC contributed substantially to the KNN (13.3 %), SVR (13.6 %), and ANN (14.3 %) models but was the least important variable in the GBoost and RFR models. In the CART model, several variables—YC, ST, EWi, EL, YH, and AH—had an importance of 0 %. Albumen height was the least important variable in the ANN, SVR, and KNN models. Discussion Descriptive statistics and correlation coefficients The investigated traits exhibited varying degrees of variability. The low variability observed in YH, EL, and EWi indicates consistent measurements within the study population. These traits are critical determinants of egg shape, which directly influences packing and transportation efficiency ( Monira et al., 2003 ). The moderate variability in EW, AH, AW, and YW may reflect different growth stages, yet it is advantageous for maintaining consistent egg size, a key quality parameter. In contrast, higher variability was observed in SW, ST, and YC. Variability in ST is significant as it directly impacts egg quality and breakage during transport. The variability in YC suggests considerable diversity, likely influenced by genetic factors, and presents an opportunity for feed modification to meet consumer preferences ( Wen et al., 2019 ). These findings align with ( Wondmeneh, 2015 ), who documented breed-specific effects on egg quality. Understanding these variabilities is essential for producers and policymakers to refine breeding programs, such as by selecting for more uniform shell thickness to reduce breakage and improve marketability ( Roberts, 2004 ; Singh et al., 2020 ). Correlation analysis revealed a strong positive relationship between EW and AW, indicating that AW is a primary driver of overall egg mass. This finding provides valuable insight for the egg industry and consumers ( Roberts, 2004 ). A substantial association was also found between EW and EL, which has direct implications for packaging and transportation dimensions ( Monira et al., 2003 ). Moderate correlations of EW with YW, EWi, and SW suggest that heavier eggs positively influence yolk mass and dimensional attributes ( Monira et al., 2003 ; Wolanski et al., 2007 ). In contrast, YH and ST showed only weak correlations with EW. The weak positive correlation between AH and EW indicates a minor influence on perceived freshness. The very weak correlation between YC and EW suggests that yolk pigmentation is influenced more by nutrition and other factors than by egg weight ( Karoui, 2025 ). The overall variability in these traits offers opportunities for targeted genetic improvement to enhance egg quality and consumer satisfaction ( Monira et al., 2003 ). Predictive performance of the algorithms The CART algorithm identified AW as the primary predictor of EW, consistent with the findings of ( Kebede & Ashenafi, 2024 ). The regression tree, originating from a root node with a mean EW of 63.46 g, used successive splits based on AW, YW, and SW to form homogeneous subgroups. The variable importance analysis confirmed that all other traits contributed 0 % to the CART model, underscoring the dominance of these three predictors. This aligns with other studies that identified AW and YW as the most significant factors for predicting EW ( Bayram et al., 2015 ; Topal et al., 2010 ). However, the primary predictive trait can vary by species; for instance, EWi was most influential in ducks ( Dahloum et al., 2024 ), and EWi was the key predictor for Nigerian helmeted guinea fowls ( Jegede et al., 2024 ). The strong performance of AW and YW in our CART model is logical, as these components constitute a major portion of egg mass ( Bayram et al., 2015 ). The terminal nodes of the tree showed a clear trend, with mean EW increasing from 57.09 g to 75.32 g as AW increased, highlighting that higher albumen mass is directly associated with greater egg weight. The GBoost model demonstrated superior predictive accuracy for EW among all algorithms tested. It achieved the highest R² and r alongside the lowest error metrics (RMSE, MAE, MSE), AIC, and BIC. This indicates an optimal balance between model complexity and predictive power. The strong performance was expected, given that the predictor variables have well-established relationships with EW, as previously demonstrated in linear models ( Bozeman et al., 2012 ). The superiority of GBoost in this context is consistent with its application in other agricultural predictions, such as sheep body condition grading ( Semakula et al., 2021 ), weaning weight in animals ( Eroğlu et al., 2025 ), and sheep body weight ( Hamadani & Ganai, 2023 ). These collective findings underscore that algorithm efficacy is highly context-dependent, varying by breed, species, and dataset characteristics ( Tırınk et al., 2023 b). RFR was the second-best performing model, surpassing ANN, KNN, SVR, and CART. Its robust performance is consistent with studies ranking RFR among the top algorithms for predicting sheep body weight ( Kozaklı et al., 2024 ) and its superiority over traditional stepwise regression in cattle ( Praharani et al., 2024 ). SVR also maintained a respectable level of accuracy in our study, aligning with its ranking as a strong performer by Kozaklı et al. (2024) . ANN and KNN demonstrated weaker performance, with ANN slightly outperforming KNN. The higher RMSE of KNN indicates a larger average prediction error. While KNN can struggle with complex, non-linear relationships in agricultural data ( Ahmad & Mariano, 2006 ), ANNs have proven effective in other animal science applications, such as estimating amino acids in feed ( Cravener & Roush, 1999 ) and clustering livestock production data ( Fernandez et al., 2006 ). CART showed the weakest performance in this study, with the highest error metrics and lowest R². This suggests a limited ability to accurately predict EW compared to the ensemble methods, potentially due to overfitting or an inability to capture complex variable interactions. It is important to note that CART remains a valuable non-parametric tool for visualizing population structures and handling outliers , and its performance is context-specific. For example, CART has outperformed other algorithms in predicting body weight in goats and poultry ( Mokoena et al., 2022 ; Tyasi et al., 2021 ), while being outperformed by MARS in cattle ( Bila et al., 2023 ; Hlokoe et al., 2022 ). This further reinforces the notion that the optimal machine learning model depends heavily on the specific dataset and research question. Conclusion The present results demonstrate that machine learning techniques provide a robust and effective alternative to conventional methods for estimating EW. A key finding was the consistent identification of AW as the most significant predictor across all models, underscoring its primary influence on overall egg mass. However, the secondary influential factors varied: YW was the second-most important variable in the RFR, GBoost, and CART models, whereas EL was secondary in the ANN, SVR, and KNN models. In terms of predictive accuracy, the GBoost algorithm was superior, achieving the highest R² of 0.960 and the lowest RMSE of 1.112. It was closely followed by RFR, which also demonstrated strong performance. The ANN, SVR, and KNN models showed moderate accuracy, while the CART model performed the weakest. This study represents the first comprehensive comparison of these algorithms for EW prediction in Ethiopia using commercial consumption eggs (table eggs). Based on the results, I recommend the GBoost algorithm as the most reliable tool for analyzing egg quality traits and estimating EW, with RFR as a viable alternative. The application of these models offers a practical and transparent approach to support breeding programs, optimize resource allocation, enhance productivity, and inform policy decisions within the poultry sector therefore; different stakeholders and policymakers may work on and improve egg quality attributes that are appropriate for customers and chick production due to the relationship between egg weight and egg parameters The study's single-breed design and lack of environmental and management data represent its primary limitations, constraining the generalizability of the findings. To address these gaps, future research should examine a wider genetic foundation and specifically assess how husbandry practices affect egg quality characteristics and their correlation with egg weight. Furthermore , validating these relationships with larger and more varied datasets—encompassing diverse breeds, age groups, and production systems—is essential. This comprehensive approach is crucial for developing robust, universally applicable strategies to optimize egg weight and quality in commercial production. Ethical statement This study was conducted in compliance with all relevant ethical guidelines for animal research. CRediT authorship contribution statement Ashenafi Getachew Megersa: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Formal analysis, Data curation, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. References Ahmad H.A., Mariano M. Comparison of forecasting methodologies using egg price as a test case. Poultry Science. 2006;85(4):798–807. doi: 10.1093/ps/85.4.789. [ DOI ] [ PubMed ] [ Google Scholar ] Akyurek H., Okur A.A. Effect of storage time, temperature and hen age on egg quality in free-range layer hens. Journal of Animal and Veterinary Advances. 2009;8(10):1953–1958. [ Google Scholar ] Kebede K., Megersa A.G. Investigation of egg quality traits affecting egg weight of Fayoumi Hen with data mining method. Journal of Veterinary Medicine and Science. 2024;1(1) doi: 10.1017/S0043933920000581. [ DOI ] [ Google Scholar ] Ali M., Eyduran E., Tariq M.M., Tirink C., Abbas F., Bajwa M.A., Jan S. 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. Pakistan Journal of Zoology. 2015;47(6) 0030-9923/2015/0006-1579 $ 8.00/0. [ Google Scholar ] Altmann A., Toloşi L., Sander O., Lengauer T. Permutation importance: A corrected feature importance measure. Bioinformatics (Oxford, England) 2010;26(10):1340–1347. doi: 10.1093/bioinformatics/btq134. [ DOI ] [ PubMed ] [ Google Scholar ] Baishya, D., Dutta, K.K., Mahanta, J.D., & Borpujari, R.N. (2008). Studies on certain qualities of different sources of chicken eggs. http://www.tanuvas.tn.nic.in/tnjvas/vol4(4)/139-141.pdf . Bayram B., Yanar M., Tüzemen N., Şahin A. Using classification and regression tree method for predicting birth weight in brown swiss cattle. Indian Journal of Animal Research. 2015;49(4):541–545. doi: 10.5958/0976-0555.2015.00108.0. [ DOI ] [ Google Scholar ] Biazen A. Doctoral Dissertation, Haramaya University; Haramaya, Ethiopia: 2021. Comparative performance evaluation of local and tropical adapted exotic breeds of chicken in Ethiopia. [ Google Scholar ] Bila L., Malatji D.P., Tyasi T.L. Predicting body weight of South African Sussex cattle at weaning using multivariate adaptive regression splines and classification and regression tree data mining algorithms. Journal of Applied Animal Research. 2023;51(1):608–615. doi: 10.1080/09712119.2023.2258976. [ DOI ] [ Google Scholar ] Bozeman S.R., Hoaglin D.C., Burton T.M., Pashos C.L., Ben-Joseph R.H., Hollenbeak C.S. Predicting waist circumference from body mass index. BMC Medical Research Methodology. 2012;12(1):115. doi: 10.1186/1471-2288-12-115. http://www.biomedcentral.com/1471-2288/12/115 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Brasil Y.L., Cruz-Tirado J.P., Barbin D.F. Fast online estimation of quail eggs freshness using portable NIR spectrometer and machine learning. Food Control. 2022;131 [ Google Scholar ] Breiman L. Random forests. Machine Learning. 2001;45(1):5–32. [ Google Scholar ] Breiman L., Friedman J.H., Olshen R.A., Stone C.J. Chapman and Hall; 1984. Classification and regression trees. [ Google Scholar ] Canaza-Cayo A.W., Churata-Huacani R., Çakmakçı C., Rodríguez-Huanca F.H., de Sousa Bueno Filho J.S., Fernandes T.J., De La Cruz Y.C.R. Use of machine learning approaches for body weight prediction in Peruvian corriedale sheep. Smart Agricultural Technology. 2024;7 doi: 10.1016/j.atech.2024.100419. Article. [ DOI ] [ Google Scholar ] Çelik Ş., Şengül T., Inci H., Söğüt B., Şengül A., Kuzu C., Ayaşan T. Estimation of egg weight from some external and internal quality characteristics in quail by using various data mining algorithms. Indian Journal of Animal Sciences. 2017;87(12):1524–1530. [ Google Scholar ] Çelik Ş., Söğüt B., Şengül T., Eyduran E., Şengül A.Y. Usability of CART algorithm for determining egg quality characteristics influencing fertility in the eggs of Japanese quail. RevistaBrasileira de Zootecnia. 2016;45(11):645–649. doi: 10.1590/s1806-92902016001100001. [ DOI ] [ Google Scholar ] Chen J.T., He P.G., Jiang J.S., Yang Y.F., Wang S.Y., Pan C.H.…Pan J.M. In vivo prediction of abdominal fat and breast muscle in broiler chicken using live body measurements based on machine learning. Poultry Science. 2023;102(1) doi: 10.1016/j.psj.2022.102239. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Cover T., Hart P. Nearest neighbor pattern classification. IEEE Transactions on Information Theory. 1967;13(1):21–27. [ Google Scholar ] Cravener T.L., Roush W.B. Improving neural network prediction of amino acid levels in feed ingredients. Poultry Science. 1999;78(7):983–991. doi: 10.1093/ps/78.7.983. [ DOI ] [ PubMed ] [ Google Scholar ] Dahloum L., Benameur Q., Yakubu A. Exploring data mining algorithms for predicting duck egg weight based on egg quality characteristics. JAPS: Journal of Animal & Plant Sciences. 2024;34(2) doi: 10.36899/JAPS.2024.2.0721. [ DOI ] [ Google Scholar ] Danil M., Efendi S., WidiaSembiring R. Journal of Physics: Conference Series. Vol. 1424. IOP Publishing; 2019. The analysis of attribution reduction of K-Nearest Neighbor (KNN) algorithm by using chi-Square. [ DOI ] [ Google Scholar ] Edriss M.A., Hosseinnia P., Edrisi M., Rahmani H.R., Nilforooshan M.A. Prediction of second parity milk performance of dairy cows from first parity information using artificial neural network and multiple linear regression methods. Asian Journal of Animal and Veterinary Advances. 2008;3(4) doi: 10.3923/ajava.2008.222.229. [ DOI ] [ Google Scholar ] Eroğlu M., Turgut A.O., Küçük M., Önen M.F. Predicting weaning weight of Romanov lambs from biometric measurements before weaning age using machine learning algorithms. Veterinary Medicine and Science. 2025;11(4) doi: 10.1002/vms3.70420. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Farooq, M., Mian, M.A., Murad Ali, M.A., Durrani, F.R., Asghar, A., & Murqarrab, A.K. (2001). Egg traits of Fayumi birds under subtropical conditions. https://www.cabidigitallibrary.org/doi/full/10.5555/20013160564 . Fernández C., Soria E., Martin J.D., Serrano A.J. Neural networks for animal science applications: Two case studies. Expert Systems with Applications. 2006;31(2):444–450. [ Google Scholar ] Freeman E.A., Moisen G.G., Coulston J.W., Wilson B.T. Random forests and stochastic gradient boosting for predicting tree canopy cover: Comparing tuning processes and model performance. Canadian Journal of Forest Research. 2016;46(3):323–339. doi: 10.1139/cjfr-2014-0562. [ DOI ] [ Google Scholar ] Friedman J.H. Greedy function approximation: A gradient boosting machine. Annals of Statistics. 2001:1189–1232. [ Google Scholar ] Gammermann A. Support vector machine learning algorithm and transduction. Computational Statistics. 2000;15(1):31–39. [ Google Scholar ] González-Recio O., Jiménez-Montero J.A., Alenda R. The gradient boosting algorithm and random boosting for genome-assisted evaluation in large data sets. Journal of Dairy Science. 2013;96(1):614–624. doi: 10.3168/jds.2012-5630. [ DOI ] [ PubMed ] [ Google Scholar ] Hamadani A., Ganai N.A. Artificial intelligence algorithm comparison and ranking for weight prediction in sheep. Scientific Reports. 2023;13(1) doi: 10.1038/s41598-023-40528-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] He, Q. (2025). A unified metric architecture for ai infrastructure: A cross-layer taxonomy integrating performance, efficiency, and cost. arXiv preprint arXiv:2511.21772. Hlokoe V.R., Mokoena K., Tyasi T.L. Using multivariate adaptive regression splines and classification and regression tree data mining algorithms to predict body weight of Nguni cows. Journal of Applied Animal Research. 2022;50(1):534–539. doi: 10.1080/09712119.2022.2110498. [ DOI ] [ Google Scholar ] Huang J.C., Tsai Y.C., Wu P.Y., Lien Y.H., Chien C.Y., Kuo C.F.…Kuo C.H. Predictive modeling of blood pressure during hemodialysis: A comparison of linear model, random forest, support vector regression, XGBoost, LASSO regression and ensemble method. Computer Methods and Programs in Biomedicine. 2020;195 doi: 10.1016/j.cmpb.2020.105536. [ DOI ] [ PubMed ] [ Google Scholar ] Jegede P., Yakubu A., Musa I.S., Vincent S.T., Shoyombo A.J., Alabi O.O.…Popoola M.A. Fertility, hatchability, and prediction of egg weight from egg quality indices of Nigerian indigenous and exotic helmeted guinea fowls. Poultry. 2024;4(1):1. doi: 10.3390/poultry4010001. [ DOI ] [ Google Scholar ] Jun M.J. A comparison of a gradient boosting decision tree, random forests, and artificial neural networks to model urban land use changes: The case of the Seoul metropolitan area. International Journal of Geographical Information Science. 2021;35(11):2149–2167. doi: 10.1080/13658816.2021.1887490. [ DOI ] [ Google Scholar ] Karabacak A., Celik S., Tatliyer A., Keskin I., Erturk Y.E., Eyduran E.…Tariq M.M. Estimation of cold carcass weight and body weight from several body measurements in sheep through various data mining algorithms. Pakistan Journal of Zoology. 2017;49(5):1731–1738. doi: 10.17582/journal.pjz/2017.49.5.1731.1738. [ DOI ] [ Google Scholar ] Karoui R. Application of fluorescence spectroscopy in egg and egg products. Application of Fluorescence Spectroscopy in Food Quality and Control. 2025;111 doi: 10.1007/978-3-031-89681-1_6. [ DOI ] [ Google Scholar ] Kecman V. MIT press; 2001. Learning and soft computing: Support vector machines, neural networks, and fuzzy logic models. [ Google Scholar ] Kozaklı Ö., Ceyhan A., Noyan M. Comparison of machine learning algorithms and multiple linear regression for live weight estimation of Akkaraman lambs. Tropical Animal Health and Production. 2024;56(7):250. doi: 10.1007/s11250-024-04049-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Mikail N., Bakir G. Regression tree analysis of factors affecting first lactation milk yield of dairy cattle. Applied Ecology & Environmental Research. 2019;17(2) doi: 10.15666/aeer/1702_52935303. [ DOI ] [ Google Scholar ] Mittal G.S., Zhang J. Prediction of temperature and moisture content of frankfurters during thermal processing using neural network. Meat Science. 2000;55(1):13–24. doi: 10.1016/s0309-1740(99)00120-5. [ DOI ] [ PubMed ] [ Google Scholar ] Mokoena K., Molabe K.M., Sekgota M.C., Tyasi T.L. Predicting body weight of Kalahari red goats from linear body measurements using data mining algorithms. Veterinary World. 2022;15(7):1719. doi: 10.14202/vetworld.2022.1719-1726. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Monira K.N., Salahuddin M., Miah G.J.I.J.P.S. Effect of breed and holding period on egg quality characteristics of chicken. International Journal of Poultry Science. 2003;2(4):261–263. [ Google Scholar ] Mummed Y.Y., Getachew A., Kebede K., Urge M. Comparative evaluation of egg quality traits in two exotic chicken breeds. J Food ChemNanotechnol. 2024;10(1):11–15. [ Google Scholar ] Olfaz M., Tirink C., Önder H. Use of CART and CHAID algorithms in Karayaka sheep breeding. Kafkas Universitesi VeterinerFakultesi Dergisi. 2019;25(1) doi: 10.9775/kvfd.2018.20388. [ DOI ] [ Google Scholar ] Ouadah A., Zemmouchi-Ghomari L., Salhi N. Selecting an appropriate supervised machine learning algorithm for predictive maintenance. The International Journal of Advanced Manufacturing Technology. 2022;119(7):4277–4301. doi: 10.1007/s00170-021-08551-9. [ DOI ] [ Google Scholar ] Peng X., Xu D. Projection support vector regression algorithms for data regression. Knowl-Based Syst. 2016;112:54–66. [ Google Scholar ] Praharani L., Talib C., Kusumaningrum D.A., Widiawati Y., Asmarasari S.A., Rusdiana S.…Saputra F. Body weight prediction of Belgian Blue crossbred using random forest. Journal of Advanced Veterinary and Animal Research. 2024;11(1):181. doi: 10.5455/javar.2024.k763. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Réhault-Godbert S., Guyot N., Nys Y. The golden egg: Nutritional value, bioactivities, and emerging benefits for human health. Nutrients. 2019;11(3):684. doi: 10.3390/null030684. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Reis I., Baron D., Shahaf S. Probabilistic random forest: A machine learning algorithm for noisy data sets. The Astronomical Journal. 2018;157(1):16. doi: 10.3847/1538-3881/aaf101. [ DOI ] [ Google Scholar ] Roberts J.R. Factors affecting egg internal quality and egg shell quality in laying hens. The Journal of Poultry Science. 2004;41(3):161–177. doi: 10.2141/jpsa.41.161. [ DOI ] [ Google Scholar ] Roush W.B., Cravener T.L., Kirby Y.K., Wideman R.F., Jr Probabilistic neural network prediction of ascites in broilers based on minimally invasive physiological factors. Poultry Science. 1997;76(11):1513–1516. doi: 10.1093/ps/76.11.1513. [ DOI ] [ PubMed ] [ Google Scholar ] SAS Institute Inc. SAS Institute Inc; 2023. JMP 17 pro [Computer software]. version 17.0. [ Google Scholar ] Semakula J., Corner-Thomas R.A., Morris S.T., Blair H.T., Kenyon P.R. Application of machine learning algorithms to predict body condition score from liveweight records of mature romney ewes. Agriculture. 2021;11(2):162. doi: 10.3390/agriculture11020162. [ DOI ] [ Google Scholar ] Shahinfar S., Kelman K., Kahn L. Prediction of sheep carcass traits from early-life records using machine learning. Computers and Electronics in Agriculture. 2019;156:159–177. [ Google Scholar ] Singh S.I.M.R.A.N., Taggar R.K., Chakraborty D.I.B.Y.E.N.D.U., Kumar A.R.V.I.N.D., Kumar N.I.S.H.A.N.T., Kumar D.H.I.R.E.N.D.R.A. Evaluation of egg quality traits of indigenous chicken of India. Haryana Veterinarian Journal. 2020;59(2):157–159. [ Google Scholar ] Sökün H., Kalkan H., Cetişli B. 2012 20th Signal Processing and Communications Applications Conference (SIU) IEEE; 2012. Classification of physical activities using accelerometer signals; pp. 1–4. [ DOI ] [ Google Scholar ] Stadelman W.J. Egg science and technology. CRC Press; 2017. Quality identification of shell eggs; pp. 39–66. [ Google Scholar ] Tırınk 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. doi: 10.3390/ani13050798. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Topal M., Aksakal V., Bayram B., Yaganoglu A.M. An analysis of the factors affecting birth weight and actual milk yield in Swedish red cattle using regression tree analysis. Journal of Animal and Plant Sciences-Japs. 2010 doi: 10.3923/javaa.2010.285.291. [ DOI ] [ Google Scholar ] Tyasi T.L., Eyduran E., Celik S. Comparison of tree-based regression tree methods for predicting live body weight from morphological traits in Hy-line silver brown commercial layer and indigenous Potchefstroom Koekoek breeds raised in South Africa. Tropical Animal Health and Production. 2021;53(1):7. doi: 10.1007/s11250-020-02443-y. [ DOI ] [ PubMed ] [ Google Scholar ] Tyasi T.L., Makgowo K.M., Mokoena K., Rashijane L.T., Mathapo M.C., Danguru L.W.…Gxasheka M. Classification and regression tree (CRT) analysis to predict body weight of Potchefstroom koekoek laying hens. Adv. Anim. Vet. Sci. 2020;8(4):354–359. doi: 10.17582/journal.aavs/2020/8.4.354.359. [ DOI ] [ Google Scholar ] Wen C., Gu Y., Tao Z., Cheng Z., Wang T., Zhou Y. Carotenoid accumulation and its contributions to yolk color. Poultry Science. 2019;98(12):6616–6623. doi: 10.3382/ps/pez478. [ DOI ] [ Google Scholar ] Wilson H.R., Suarez M.E. The use of egg weight and chick weight coefficients of variation as quality indicators in hatchery management. Journal of Applied Poultry Research. 1993;2(3):227–231. doi: 10.1093/japr/2.3.227. [ DOI ] [ Google Scholar ] Wolanski N.J., Renema R.A., Robinson F.E., Carney V.L., Fancher B.I. Production, modeling, and education. Poultry Science. 2007;86:1784–1792. doi: 10.1093/ps/86.8.1784. [ DOI ] [ PubMed ] [ Google Scholar ] Wondmeneh Esatu Woldegiorgiss, W.E.W. (2015). Genetic improvement in indigenous chicken of Ethiopia. [ http://edepot.wur.nl/343608 ]. Wu X., Zhang Y., Shi M., Li P., Li R., Xiong N.N. An adaptive federated learning scheme with differential privacy preserving. Future Generation Computer Systems. 2022;127:362–372. [ Google Scholar ] Yang, Y., Mekki, D.M., Lv, S.J., Yu, J.H., Wang, L.Y., Wang, J.Y., … & Dai, G.J. (2006). Canonical correlation analysis of body weight, body measurement and carcass characteristics of Jinghai Yellow chicken. http://198.170.104.138/medwellonline/java/2006/980-984.pdf . Articles from Veterinary and Animal Science are provided here courtesy of Elsevier ACTIONS View on publisher site PDF (1.2 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