Occupancy Detection in Smart Buildings Using Support Vector Regression Method

This paper proposes a novel approach which applies support vector regression (SVR) method to detect office occupancy in smart buildings. The model is able to detect occupancy with two sets of features corresponding to different degrees of convenience. The feature pool is comprised of solar factor, working time, lights energy, indoor temperature, and outdoor temperature, all of which are obtained through mathematical simulation based on the thermal software EnergyPlus. We show different results by setting different parameter values in the SVR model, and demonstrate model comparison on occupant detection. The proposed model displays advantages in occupancy detection which is capable of making accurate detections at low requirements for sensor accuracy and reducing the equipment cost in smart buildings.