EUVIMEDEUVIMED entdecken
Crossref

Predicting dry matter intake in dairy cattle using machine learning models to enhance precision on dairy farms

Edwin Jun Kiat Ong, Conrad P. Ferris, Masoud Shirali

CABI Agriculture and Bioscience · 05.05.2026

DOI: 10.1079/ab.2026.0039 ↗

Kurz gefasst

Fachartikel in der Zeitschrift CABI Agriculture and Bioscience, online erschienen am 05.05.2026. Verfasst von Edwin Jun Kiat Ong und 2 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Within the context of a sustainable dairy industry, optimizing feed resource utilization is crucial. Nevertheless, the growing availability of data from a wide range of on-farm technologies may create opportunities to improve dry matter intake (DMI) predictions, especially if we can leverage machine learning (ML) algorithms. This study aims to predict DMI in Holstein dairy cows by applying both linear and non-linear ML methods to data which is, or can be, routinely available on commercial farms, including milk production and herd management data.

Methods

Data from five experiments were utilized, all of which involved feeding concentrates based on a feed-to-yield basis. Weekly records from four experiments (energy corrected milk yield, live weight, body condition score, lactation number, weeks-in-milk and milk fat-to-protein ratio) were used to train linear and non-linear models to predict DMI. Data from the fifth experiment was used as an independent validation set. Wilcoxon signed-rank test was applied on the model performance metrics to evaluate the predictive performance.

Results

From the different modelling approaches, the random forest algorithm provided the lowest errors with the highest coefficient of determination, indicating it has the best predictive performance. Energy corrected milk yield was identified as a key parameter in predicting feed intake, having the greatest variable importance score in the random forest model. Overall, the model predicted DMI with a moderate degree of accuracy, despite observed inaccuracies at the start and 20 weeks post-lactation.

Conclusions

This study serves to demonstrate the potential of ML models to predict DMI using a limited number of readily available on-farm parameters. The findings suggest that expanding the dataset to include a more diverse range of data would be crucial to develop a more reliable and generalizable tool for precision feed management.

Zitieren

APA

Ong, E. J. K., Ferris, C. P., & Shirali, M. (2026). Predicting dry matter intake in dairy cattle using machine learning models to enhance precision on dairy farms. CABI Agriculture and Bioscience. https://doi.org/10.1079/ab.2026.0039

Vancouver

Ong EJK, Ferris CP, Shirali M. Predicting dry matter intake in dairy cattle using machine learning models to enhance precision on dairy farms. CABI Agriculture and Bioscience. 2026. doi:10.1079/ab.2026.0039

BibTeX
@article{Ong2026,
  author = {Ong, Edwin Jun Kiat and Ferris, Conrad P. and Shirali, Masoud},
  title = {Predicting dry matter intake in dairy cattle using machine learning models to enhance precision on dairy farms},
  journal = {CABI Agriculture and Bioscience},
  year = {2026},
  doi = {10.1079/ab.2026.0039}
}
Kostenlos starten →