An adaptive framework for applying machine learning in smart spaces

A smart space ¹ is a physical environment that contains cooperating nodes which continuously and autonomously monitor their surroundings. The environment can interact with users and adapt their behaviors to enhance user experiences using semantic reasoning. Such semantic reasoning is based on information gathered and shared either from the physical environment (e.g., via sensors) or from the Internet (e.g., via user profiles). The nodes share knowledge to adapt their behaviors using semantic reasoning in a smart space. On the other hand, machine learning is a promising tool to generate or enhance knowledge for nodes' adaptations. In this paper, we propose a semantic learning component in a comprehensive smart space architecture to generate knowledge on stored semantics for nodes' adaptations. For this purpose, we propose an adaptive framework which includes machine learning techniques in the component for nodes' behaviors. Moreover, an example use-case is presented using the K-nearest neighbor algorithm. Further, two use-cases are discussed in support of the proposed framework. Finally, we address the further work to be studied.