MOTION AND ACTIVITY-DRIVEN ENSEMBLE MODELS FOR CLASSIFYING BROILER HEALTH STATUS

Abstract

Chicken production industry is a multi- billion dollars investment and a major economic sector in Nigeria. Though a viable industry, poultry(broiler) production has a major challenge of disease outbreaks that can threaten large investment into the business and can lead to high level of mortality. However, early detection of poultry health in a large and commercial quantity is difficult. Artificial intelligence and machine learning have become important components in health management of poultry farming. The use of AI and other technology like computer vision are used to monitor chicken health. As a machine learning-based health detection solution, this paper developed ensemble models using Random Forest, Adaboost and bagging for classification of health status of broilers based on their motion/physical activity. The dataset was acquired from an experimental work that track and monitored the motions of healthy and sick birds in a farm. The result of the model shows that all the 3 models achieved accuracy level of 97.9%. The models can be implemented in a system that monitored broilers and classify the health status, this will lead to early detection and reduce mortality.