This research paper presents a Linear Regressor (LR) modeling approach for automated thermal conditioning of smart spaces in a classroom environment using limited data constraints. Sensitive studies were performed in order to identify the breaking point of the LR model of polynomial-order-of-1 below which it will not be possible to obtain meaningful estimates and controllability actions, considering a range between 5% and 30% limited training data points. Based on several simulation experiments, it was found that the LR breaks at the 10% training data level. This result is valuable for recommender systems that may experience difficulty in obtaining enough primary dataset. Applications that provide early warning signals will also find this study useful.
Keywords: linear regressor, primary dataset., recommender systems, simulation experiment, smart classroom
