Machine learning, indoor air temperature
This paper investigates a mechanism in which indoor air temperature can be predicted using the temperature sensor on the battery and the CPU of smartphones or laptops, where data is easily accessible and ubiquitous. As a case study, several machine-learning methods were used to build models from a MacBook Pro’s data and the measured surrounding air temperature. The effects of the machine learning type and input feature size (by including other parameters such as CPU processing usage and battery charge percentage) on the model accuracy were investigated. The goal is to determine a set of feature combinations that can be used to build models which can accurately predict the indoor temperature. The accuracy of these models was measured by comparing their prediction to the actual indoor temperature.