Multi-Sensor Integration and Machine Learning for High-Resolution Classification of Herbivore Foraging Behavior

Source
DataCite
Published
2025
License
Creative Commons Zero v1.0 Universal
Themes
Animal Welfare, Agriculture & Farming, Machine Learning & AI

Get the data from DataCite

About this dataset

The study used Random Test-Split (RTS) and Cross-Validation (CV) machine learning methods to test different models to classify cattle behavior foraging behaviors states, foraging activities, posture, and activity by posture, using GPS coupled accelerometer data with 12-hour / days continuous recording observation as supporting ground truth. RTS in XGBoost performing best for general activity state classification, while CV in Random Forest excelled in more detailed foraging activities and activity-posture classifications. Key movement indicators like speed, Actindex and sensor values (x, y, and z) were vital in predicting behaviors, suggesting specific sensors for tracking behaviors of interest to ranchers. The results highlight the benefits of continuous monitoring and advanced data analysis for real-time livestock tracking, leading to better grazing management, improved animal...

How the data was collected

Key movement indicators like speed, Actindex and sensor values (x, y, and z) were vital in predicting behaviors, suggesting specific sensors for tracking behaviors of interest to ranchers.

Browse sustainability datasets on SusanHub