Machine Learning Model as a Service in Smart Agriculture Systems

Machine learning applications that rely on trained machine learning models are a major area of interest within the field of Smart Agriculture Systems. One of the greatest challenges in this area is to ensure that the machine learning model is up-to-date and easily and effectively deployed to all smart systems that rely on it. However, the rapid changes, or drifts in the data can have a serious effect on the accuracy of the model, leading to unforeseeable consequences in system behavior. Whilst some research has been carried out on ML model updates, this is still an open challenge in modern IoT-based Smart Systems. This paper reviews different ways in which ML models can be served and updated, and proposes a detailed architecture for Edge part of Edge-Cloud-based Smart Agriculture System. To support this architecture, a prototype system that provides a machine learning model as a service in the agricultural domain is developed and tested.