A federated learning model for intelligent cattle health monitoring system using body area sensors and IoT

Abstract

The Sustainable Development Goals (SDGs) emphasize synchronizing technology and routine life for sustainability. Food and water shortage, and exponentially increasing environmental pollution are the biggest challenges for sustainability. Livestock plays a vital role in developing countries’ economies; the most profitable businesses are breeding dairy and non-dairy products. The productivity of cattle farms is dependent on the health conditions of cattle. Identifying unhealthy cattle and providing suitable treatment is critical. Hence, deploying the Internet of Things (IoT) along with AI systems is one of the potential solutions. This cattle health monitoring system provides monitoring of cattle health to ensure the minimum human intervention. A system has been designed and developed to aid the intelligent cattle health monitoring system by using machine learning techniques. The system includes multiple sensor nodes, each having a body area sensor that is connected to the IoT platform through a controller. As a novelty, the prototype has been trained and evaluated using a federated learning technique. The system warns the owner about specific diseases such as fever, mastitis, foot and mouth disease, and ketosis. The presented results validate the proposal as it diagnoses the prescribed viral diseases precisely. We have implemented the Gaussian Naïve Bayes classifier for this multiclass problem. Considering the federated learning model, three different datasets are considered as three different clients with 70% train and 30% test data. Client 1, Client 2, and Client 3 represent the cattle farm, veterinary hospital, and veterinary respectively. The sensor nodes are placed on key points of the cattle body while each node collects physiological parameters that are further used to train the prediction system. Additionally, we have developed a user-friendly Android application for the owner to control cattle well-being. A comprehensive comparative analysis demonstrates that the proposed system outperforms existing state-of-the-art systems by showing good accuracy.