Reducing the ecological footprint of animal agriculture is crucial for sustainable precision agriculture. Accurately predicting dry matter intake (DMI) in cattle is a key strategy to achieve this goal. In this study, we utilized animal intake data with climatic data to predict the dry matter intake using advanced machine learning (ML) models including Gaussian Process Boosting (GPBoost). The full dataset contains 12,056 daily feed, water, and climatic records from 178 cattle collected at the West Virginia University dry lot during winter and spring bull/steer testing. We attempted to develop a robust ML pipeline, and our findings revealed that the GPBoost model outperformed gradient-boosting regression, light gradient-boosting regression, and Extreme gradient boosting regression with optimal balance between bias and variance. Additionally, we employed Shapley’s additive explanation (SHAP) to provide explainable Artificial intelligence insights to understand the model’s predictions. Moreover, we demonstrate that the climate variables are not necessary for the DMI prediction. This study reports predictive algorithms for application in animal agriculture and the integration of models into practical tools using ML tools (MLflow and docker) for real-time predictions and animal monitoring.

Read the full article in the Journal of Animal Science.