Indoor CO2 Level-Based Occupancy Estimation at Low-Scale Occupant using Statistical Learning Method

Abstract

Most of the occupancy estimations based on indoor CO2 levels are tested on a large-scale number of occupants such the order of tens or hundreds. Logically because the pattern of the occupancy and CO2 level is about similar at a broad range of occupants. In the present study, a small office room with an occupancy scale of 0-6 people was tested. The Statistical Learning method is used to estimate the number of occupants, including Decision Tree, Random Forest classifier, SVM, Logistic regression, K-Nearest Neighbor, and Neural Network. A combination of training and testing data set is applied to the methods and a comparison has been made in order to distinguish their accuracy. The result shows that the accuracy of self-estimation and cross-estimation is ranged from 86-100% and 86-94% respectively. It also found that the estimation accuracy of self-and-cross validation does not significantly increase with the increase of data set combination.