Evaluating machine learning models for optimal livestock environment prediction in smart farming applications to enhance food security

Livestock is a vital protein source for the global population and any disruptions of supply can significantly threaten national food security. Therefore, ensuring a stable and continuous livestock supply is essential. Researches have highlighted a strong link between livestock production and environmental factors and proposed several smart farming solutions to be adopted to effectively monitored and optimized livestock production. In this study, we propose the adoption of sensor technology and machine learning to establish optimal environmental conditions for livestock in Malaysia. Evaluation of various machine learning techniques is carried out to determine the most effective model for enhancing livestock production in the smart farming systems. This research simulates the livestock living environment equipped with sensors and selected parameters for data collection to be utilised in training the selected machine learning models which are Decision Tree, Naive Bayes and K-Nearest Neighbours. The trained machine learning models are then applied in predicting the optimum environment for livestock using the dataset of simulated environment. Then, performance evaluation on the machine learning models was carried out. The accuracy results for Decision Tree, Naive Bayes and K-Nearest Neighbours are 99%, 63% and 89% respectively. From the research, Decision Tree model is found to be the best performing model at predicting the optimum environment for livestock. These findings provide invaluable insight to advance researches on optimum livestock environment prediction in smart farming for Malaysia use case and will enable precise adjustments and monitoring to achieve ideal conditions for livestock growth to provide consistent livestock production.