Indoor Wellness: a Machine Learning Technique for Internet of Things Sensor Data

The rise in stress levels exerts a profound influence on both the physical and mental well-being of individuals often leading to burnout syndrome, especially for those working in high demand fields such as healthcare. Stress accumulates through various pathways and is particularly influenced by the environmental condition of the subject, as an example people living in rainy regions are more likely to suffer from depression than those living in mild climate regions. Both EN ISO 7730 and ASHRAE 55 dictate the minimal requirements and standards for a living environment to be satisfying to the users by classifying the rooms in three different classes based on the PMV (Predicted Mean Value) which quantifies the average thermal comfort of the inhabitants, measured through different physical quantities often hard to measure. The aim of the project is to combine IoT technology and machine learning approaches to address this difficulty using simpler features and devices. The data is gathered by placing two Tauanito INAR devices, curtesy of TAUA s.r.l., in three rooms of the “Alma Mater Studiorum – Cesena Campus”. The analysis are comprehensive of a decomposition one aimed at exploring the correlation between the variables, a classification one which facilitates the ranking of the rooms according to the norms, a first regression analysis for the prediction, both linear and non, of the satisfaction scores and a second regression one for the prediction of the optimal temperature for which the satisfaction is maxed. Although more studies are required to effectively validate the tasks, the results are promising in all four tasks and open up the possibility for machine learning to improve people’s life under the psycho-physical aspect.