Internet of Things Based Occupancy Detection Using Ensemble Classifier for Smart Buildings

Buildings account for a large share in energy consumption in day to day life. Occupancy based models can help in modeling whether the building or the particular room is occupied or not. Occupancy detection mechanisms can help in automating the electrical appliances and make them operational only in presence of the person in the room. This helps in creating energy aware scenarios which can contribute to the energy efficiency and reduction in power tariff. In this work, we take the approach of occupancy modeling by the help of Ensemble Models constructed using Random Forests, Logistic Regression and Support Vector Machine classifiers. The ensemble approach undertaken in this work is Voting and the weights of classifiers to the meta-model are fine-tuned using a Differential Evolution optimization algorithm. The results were found to be of high accuracy, i.e., 98.8% and 98.7% on the given test sets.