Eco Roam Poultry

Abstract

Poultry Farming is crucial to feeding the world's growing population. Birds' abnormal behaviour can harm the birds, and disease detection relies on poultry behaviour. Integrating Internet of Things (IoT) technology into poultry farming can revolutionize the way to monitor and manage poultry health. Feeding, preening, and dustbathing are poultry's daily routines. In response to the problem of detecting correct poultry behaviour and health status, this paper proposes a smart poultry monitoring system that leverages IoT sensors to detect and monitor chicken behaviour in poultry farms and provides valuable information to industry stakeholders for management decisions and individual poultry health status. The phases of the proposed system are data preprocessing, feature extraction, feature selection, and detection of poultry behaviour via different classification algorithms. An optimized synthetic minority over-sampling technique (SMOTE) via an artificial hummingbird algorithm (AHA) is applied to solve the data imbalance problem. The experimental results show that an optimized SMOTE obtains better accuracy with 97 % than other algorithms. Further, to attain accuracy in predicting poultry behaviours, Random Forest (RF) achieves superiority compared to other machine learning algorithms with an accuracy of 98 %.