Detection of Broiler Behaviors Through a Wearable Sensor System and Machine Learning Methods

The wellbeing of broiler chickens is an increasing concern in society. This has a direct bearing on the standard of animal farmed products and the overall health of those who eat them. Importantly, taking good care of animals has a big impact on helping broilers grow healthy. This also leads to fewer diseases and makes the quality and amount of bird products better. This study's primary objective is to categorize different broiler behaviors since these responses are important welfare markers. This endeavor is achieved through the meticulous analysis of data acquired from wearable accelerometers, utilizing an array of machine learning models, namely Discriminant Analysis (DA), Decision Trees (DT), Naïve Bayes (NB), K-Nearest Neighbour (KNN), Support Vector Machine (SVM), and Ensemble Classifiers. The Quadratic-SVM study showed the best performance of 84.8% and an F-measure of 86.1% overall, while the Coarse Gaussian-SVM analysis showed the lowest performance of 15.0% and an F-measure of 15.7 %. The results show our method's potential for accurately determining specific broiler behaviors, making it helpful for implementation in the future.