Thermal comfort in office buildings is emerging as an important variable that can be used to maximize employee productivity. In this paper we propose a new Internet of Things (IoT) based system that creates a personalized model of thermal comfort.
To create this model, our system collects telemetry via an IoT network of sensors and user inputs. This data is then input into machine learning algorithms that continuously calibrate and update a personalized thermal comfort model for the user. To facilitate the individuality of our models, the system combines personal measurements from the Microsoft Band, such as biometric readings and user feedback, with environmental measurements such as temperature, humidity, and air speed. In this work, we evaluate a broad set of classification and regression algorithms.
Our experimental results show that using our IoT based system improves the mean squared error of the thermal prediction by about 50% when compared to the industry standard method developed by P.O. Fanger.