An Unsupervised Rule Recommendation System in a Heterogeneous Smart Home

The field of Smart Homes and enhanced living environments is a growing industry and the need of more intelligent solutions is present. The Internet of Things generates more and more data and various devices are connected within modern homes. Using this data to improve the living quality and raise the comfort of people is an interesting question and the range of application domains include Smart Homes, Smart Cities, Smart Energy Consumption among others. This work handles the potential of structured data collection in Smart Homes to enhance the comfort of users, by recommending home automation rules. The thesis defines requirements for data collection in heterogeneous Smart Homes and propose an architecture for this purpose. Structured Smart Home datasets are used to apply state-of-the-art machine learning algorithms and propose a system for the discovery and recommendation of home automation rules. Experimental results show the relevance for home automation efforts and can be applied in real world use cases.