View in EDS

Prediction of Body Mass of Dairy Cattle Using Machine Learning Algorithms Applied to Morphological Characteristics.

Saved in:
Bibliographic Details
Title: Prediction of Body Mass of Dairy Cattle Using Machine Learning Algorithms Applied to Morphological Characteristics.
Authors: de Oliveira, Franck Morais, Ferraz, Patrícia Ferreira Ponciano, Ferraz, Gabriel Araújo e Silva, Pereira, Marcos Neves, Barbari, Matteo, Rossi, Giuseppe
Source: Animals (2076-2615); Apr2025, Vol. 15 Issue 7, p1054, 22p
Subject Terms: ARTIFICIAL neural networks, MACHINE learning, ARTIFICIAL intelligence, REGRESSION analysis, STATISTICAL correlation
Abstract: Simple Summary: Predicting body mass (BM) in dairy cattle is essential for efficient herd management, optimizing feeding strategies, and monitoring animal condition. Traditional methods, such as direct weighing, can be labor-intensive and impractical in large-scale production systems. This study explored the use of advanced computational techniques, including artificial neural networks (ANNs) and Support Vector Regression (SVR), alongside traditional regression models, to estimate the BM based on morphological data. Thoracic and abdominal perimeters were identified as highly correlated variables, enabling the development of high-accuracy predictive models. The findings highlight the potential of computational approaches to improve BM estimation, providing practical alternatives for the livestock sector. While more complex models demonstrated superior predictive performance, simpler statistical methods remain valuable options for on-farm adoption, balancing accuracy and ease of implementation. The accurate prediction of body mass (BM) in cattle is crucial for herd monitoring, assessing biological efficiency, and optimizing nutritional management. This study evaluated BM prediction models using morphological data from 465 lactating Holstein cows, including the dorsal length (DL), thoracic width (TW), abdominal width (AW), rump width (RW), hip height (HH), body depth (BD), thoracic perimeter (TP), and abdominal perimeter (AP). Spearman's correlation analysis identified TP (r = 0.89), AP (r = 0.88), and RW (r = 0.80) as the strongest predictors. Simple and multiple linear regression models, artificial neural networks (ANNs), and Support Vector Regression (SVR) were tested. The dataset was split into 90% for training (419 samples), 5% for validation (23 samples), and 5% for testing (23 samples). The best simple model, using only TP, achieved an R2 of 0.7763 and an RMSE of 43.69 kg. A multiple regression model with TP, AP, and RW improved performance (R2 = 0.9067, RMSE = 28.00 kg). The ANN outperformed all of the models (R2 = 0.9125, RMSE = 25.86 kg), and was followed by SVR (R2 = 0.9046, RMSE = 27.41 kg). As an indication of the evaluation of the results obtained, it is observed that, although regression models are effective, the ANNs and SVR provide greater accuracy, reinforcing their potential for herd management. However, simpler models remain viable alternatives for practical on-farm application. [ABSTRACT FROM AUTHOR]
Copyright of Animals (2076-2615) is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Complementary Index
Description
Abstract:Simple Summary: Predicting body mass (BM) in dairy cattle is essential for efficient herd management, optimizing feeding strategies, and monitoring animal condition. Traditional methods, such as direct weighing, can be labor-intensive and impractical in large-scale production systems. This study explored the use of advanced computational techniques, including artificial neural networks (ANNs) and Support Vector Regression (SVR), alongside traditional regression models, to estimate the BM based on morphological data. Thoracic and abdominal perimeters were identified as highly correlated variables, enabling the development of high-accuracy predictive models. The findings highlight the potential of computational approaches to improve BM estimation, providing practical alternatives for the livestock sector. While more complex models demonstrated superior predictive performance, simpler statistical methods remain valuable options for on-farm adoption, balancing accuracy and ease of implementation. The accurate prediction of body mass (BM) in cattle is crucial for herd monitoring, assessing biological efficiency, and optimizing nutritional management. This study evaluated BM prediction models using morphological data from 465 lactating Holstein cows, including the dorsal length (DL), thoracic width (TW), abdominal width (AW), rump width (RW), hip height (HH), body depth (BD), thoracic perimeter (TP), and abdominal perimeter (AP). Spearman's correlation analysis identified TP (r = 0.89), AP (r = 0.88), and RW (r = 0.80) as the strongest predictors. Simple and multiple linear regression models, artificial neural networks (ANNs), and Support Vector Regression (SVR) were tested. The dataset was split into 90% for training (419 samples), 5% for validation (23 samples), and 5% for testing (23 samples). The best simple model, using only TP, achieved an R<sup>2</sup> of 0.7763 and an RMSE of 43.69 kg. A multiple regression model with TP, AP, and RW improved performance (R<sup>2</sup> = 0.9067, RMSE = 28.00 kg). The ANN outperformed all of the models (R<sup>2</sup> = 0.9125, RMSE = 25.86 kg), and was followed by SVR (R<sup>2</sup> = 0.9046, RMSE = 27.41 kg). As an indication of the evaluation of the results obtained, it is observed that, although regression models are effective, the ANNs and SVR provide greater accuracy, reinforcing their potential for herd management. However, simpler models remain viable alternatives for practical on-farm application. [ABSTRACT FROM AUTHOR]
ISSN:20762615
DOI:10.3390/ani15071054