This project emphasizes on designing and developing a personalized service for Smart Homes and their inhabitants, with the objective of achieving a self-operated house. The main idea is that all “smart” devices installed in a house would be able to function by themselves without human intervention, based on the preferences of the house’s residents. To achieve that, several technologies need to be utilized.
Internet of Things (IoT), Cognitive Computing and the Cloud are used in order to control the “smart” devices and collect historical data over time. These data contain information about user preferences inside a house, e.g. a room’s temperature, or if a room’s lights are on. A Cloud-based IoT platform was utilized, along with several software-developed IoT devices (simulating real “smart” devices in a house), with the purpose of collecting the aforementioned historical dataset. The Smart Home user can interact with the IoT devices directly, or by using a smartphone application. The application was developed to exploit Cloud-based Cognitive Computing capabilities, resulting in a more natural interaction between the user and the IoT devices. Once a dataset has been collected, several Supervised Machine Learning algorithms are used, as well as Time Series Forecasting, in order to train Machine Learning models. These models are trained based on the aforementioned dataset. Following that, the trained models are evaluated in order to detect the one performed best. Then, the model which provided the finest results is selected to be used in a Smart Home environment.
Finally, the smartphone application can contact the previous trained Machine Learning model, through a web service, and request future predictions for the status of all IoT devices inside a house over time. Then, the status of all IoT devices is adjusted according to the received forecasts.