Machine Learning based Estimation of Room Occupancy Using Non-Intrusive Sensors

Buildings use a lot of energy to heat, ventilate, and cool. Making them demand-driven based on human occupancy is one way to improve their efficiency. Supervised learning approaches such as Multiclass Support Vector Machine (MSVM), Linear Discriminant Analysis (LDA), and Bagged Tree (BT) were used in this study to focus on a variety of different combinations of feature sets. Furthermore, we evaluated the performance of our models using a range of performance measures which, includes accuracy, specificity, precision, sensitivity, and the F-measure. The results of the trials reveal that determining the largest number of individuals in a room can be done with 99.7% accuracy, excellent sensitivity, and specificity.