Machine Learning Based Occupancy Detection via the Use of Smart Meters

In this work, the challenging problem of occupancy detection in a domestic environment is studied based on information gathered from electricity and water consumption smart meters. The most popular machine learning techniques, along with their boosting versions, are utilized for occupancy detection using the measurements of a door counter sensor as ground truth for training. In order to evaluate information gained from electricity and water consumption features and to reduce dataset sparsity, while maintaining the performance of classification techniques, mutual information is used as feature extraction technique. In order to determine the most efficient parameter combinations of machine learning techniques, we performed a series of Monte Carlo simulations for each method and for a wide range of parameters. Our simulation results show a superiority of Random Forest learning technique compared to the other classification techniques with accuracy slightly over 80% and F-measure with almost 84%, respectively.