Predicting dry matter intake in cattle at scale using gradient boosting regression techniques and Gaussian process...

Area of Science:

• Agricultural Science
• Machine Learning
• Animal Science

Background:

• Dry matter intake (DMI) is crucial for livestock management, but accurate measurement of grazing cattle is challenging.
• Existing methods for dry lot DMI quantification are expensive and not applicable to grazing systems.
• Machine learning (ML) offers a potential solution for predicting DMI, improving efficiency and enabling participation in ecosystem service programs.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting individual dry matter intake (DMI) in beef cattle.
• To compare the performance of various gradient boosting approaches against baseline models.
• To establish a scalable and reproducible ML workflow for DMI prediction.

Main Methods:

• Utilized a dataset of 12,056 daily records from 178 beef cattle at West Virginia University (2019-2020).
• Explored gradient boosting regression (GBR), LightGBM (LGB), XGBoost (XGB), and Gaussian Process Boosting (GPBoost) models.
• Developed an end-to-end MLOps pipeline using MLflow and Docker for streamlined ML operations and deployment.

Main Results:

• Gaussian Process Boosting (GPBoost) models demonstrated superior performance with the best bias and variance compared to other models.
• Optimized GPBoost models achieved RMSE values ranging from 1.18 to 1.54 kg on test data.
• GPBoost models achieved R-squared values of 0.58 (training) and 0.55 (testing), with MAE values of 0.92 kg (training) and 0.90 kg (testing), indicating good generalization.

Conclusions:

• GPBoost models provide a reliable and accurate method for predicting beef cattle DMI, outperforming traditional and other ML approaches.
• The developed MLOps pipeline ensures reproducibility, scalability, and seamless deployment of the DMI prediction model.
• Accurate DMI prediction using ML can enhance livestock management efficiency and facilitate producer participation in ecosystem service credit programs.