Health Status Classification for Cows Using Machine Learning and Data Management on AWS Cloud

Area of Science:

• Agricultural Science
• Data Science
• Veterinary Medicine

Background:

• Livestock health is crucial for agricultural sustainability and profitability.
• Efficient monitoring of individual cow health is vital for preventing disease outbreaks and maintaining herd productivity.
• Non-invasive monitoring technologies offer a promising approach to real-time health assessment.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for classifying milk cow health status into three categories.
• To establish a systematic data processing and ML modeling workflow for dairy farm applications.
• To integrate the ML model into a cloud-based architecture for real-world deployment.

Main Methods:

• Data collection from non-invasive IoT devices on dairy farms, including environmental and individual cow metrics (age, lactation status).
• Systematized data processing workflow for cleaning, feature engineering, and model training.
• Training and evaluation of five ML algorithms, with a focus on the Random Forest Classifier (RFC).
• Development of a cloud architecture on Amazon Web Services (AWS) for model deployment and a client application interface for result visualization.

Main Results:

• The Random Forest Classifier (RFC) achieved the highest performance in health status assessment.
• RFC model demonstrated high accuracy (0.959), recall (0.954), and precision (0.97).
• A comprehensive data processing workflow and cloud integration strategy were successfully implemented.

Conclusions:

• The developed ML model, particularly the RFC, effectively classifies cow health status, offering a valuable tool for dairy farm management.
• The proposed workflow and cloud architecture provide a scalable and reliable solution for real-time livestock health monitoring.
• This approach contributes to improved animal welfare, enhanced farm productivity, and increased agricultural sustainability.