Classification and Analysis of Multiple Cattle Unitary Behaviors and Movements Based on Machine Learning Methods

Area of Science:

• Animal Behavior
• Machine Learning in Agriculture
• Livestock Monitoring

Background:

• Livestock behavior is a key indicator of animal welfare and health.
• Accurate monitoring of animal behavior is crucial for farm management.
• Previous methods for behavior classification have limitations.

Purpose of the Study:

• To develop a framework using inertial measurement unit (IMU) data for classifying dairy cow behaviors.
• To identify specific movements within classified behaviors.
• To evaluate the performance of different machine learning algorithms for behavior classification.

Main Methods:

• Collected IMU data from 10 dairy cows.
• Classified six unitary behaviors: feeding, standing, lying, ruminating-standing, ruminating-lying, and walking.
• Investigated K-nearest neighbors (KNN), random forest (RF), and extreme boosting algorithm (XGBoost) performance across 5, 10, 30, and 60-second time windows.
• Analyzed acceleration data for specific feeding-related movements (feed tossing, rolling biting, chewing).

Main Results:

• The extreme boosting algorithm (XGBoost) achieved the highest classification performance.
• XGBoost demonstrated an average F1 score of 94% for the six unitary behaviors within a 60-second time window.
• Movement classification achieved F1 scores of 78% for feed tossing, 87% for rolling biting, and 87% for chewing.

Conclusions:

• The proposed framework effectively classifies dairy cow unitary behaviors using IMU data.
• XGBoost is a highly effective machine learning algorithm for this classification task.
• The framework enables detailed analysis of specific animal movements within broader behavioral categories.