Comparison of Machine Learning methods for Categorizing Objects in the Internet of Things

Objects connected to the IoT are of several types and can be categorized according to their specific capabilities. These categories are an indicator of what actions the object can perform and the extent to which each of them can be performed. For IoT systems designers, these categories serve as a guide for choosing the correct objects and devices for achieving the goals of the system within the boundaries of requirements, specifications, and constraints. For end-users, categories provide a broad view of the benefits they will get from the use of the object or system, as well as assessing the potential risks of allowing access to personal data. This paper evaluates the most comprehensive machine learning-based categorization schemes proposed for the IoT and compares the methods used to categorize objects in each of them.