Early Warning Accuracy of a Behavioral Precision Livestock Farming Platform
Year: 2026
Type: conference-paper
Abstract: Behavioral telemetry from precision livestock farming (PLF) systems offers the potential to generate actionable disease alerts several days before clinical inspection is feasible. This study reports on a production-scale deployment involving 130 Hungarian dairy cows and 12,872 behavioral observations collected between October 2025 and February 2026, complemented by high-frequency inertial sensing pilots. Using lactation-aware feature engineering and a class-balanced random forest trained to anticipate attention flags 24 h ahead, the system achieved an area under the ROC curve of 0.77, with 0.48 recall and 0.36 precision on previously unseen animals. Eating duration, rumination percentage, and seven-day behavioral trend derivatives contributed most strongly to predictive performance, and the resulting health-score tiers were aligned with farm response protocols. The study also describes the end-to-end system architecture and data quality controls, while highlighting limitations related to missing veterinary ground truth and incomplete sensor-to-animal mappings. This study presents a reproducible analytical pipeline and establishes a transparent accuracy baseline for early-warning PLF deployments.
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W7204188743