Machine learning for resource management in smart environments

Abstract

Efficient resource and energy management is a key research and business area in todays IT markets. Cyber-physical ecosystems, like smart homes (SHs) and smart Environments (SEs) get interconnected, the efficient allocation of resources will become essential. Machine Learning and Semantic Web techniques for improving resource allocation and management are the focus of our research. They allow machines to process information on all levels, inferring expressive knowledge from raw data, in particular resource predictions from usage patterns. Our aim is to devise a novel approach for a machine learning (ML) and resource Management (RM) framework in SEs. It combines ML and Semantic Web techniques and integrates user interaction The main objective is to enable the creation of platforms that decrease the overall resource consumption by learning and predicting various usage patterns, and furthermore making decisions based on user-feedback. For this purpose, we evaluate recent research and applications, elicit framework requirements, and present a framework architecture. The approach and components are assessed and a prototype implementation is described.