Sensor data fusion using machine learning techniques in indoor occupancy detection

Machine learning techniques have been considered for the fusion of environmental sensor data to detect indoor occupancy. Indoor occupancy detection is carried out to monitor and control HVAC, efficient utilization of infrastructure, to provide an effective working environment, and also maintain social distancing in the current scenario. The analysis and comparison of various ML techniques on the UCI repository occupancy dataset is cornerstone of the paper. The dataset consists of temperature, humidity, light, and CO2 sensor data. WEKA simulation tool is used to pre-process and apply classification algorithms on the downloaded occupancy dataset. Classifiers have been first applied on single sensor data and after that on the combination of sensors data to see the impact of accuracy on results. Performance of the classifiers is evaluated on metrics like TP, FP, TN, FN, Specificity, Sensitivity, Precision, FPR, RMSE, RRSE, and accuracy to detect occupancy in a lab room precisely. Future 388work will include the preparation of our environmental sensor dataset and algorithm design.