Data analytics for smart buildings: a classification method for anomaly detection for measured data

Data generated by the increasingly frequent use of sensors in housing provide the opportunity to monitor, manage and optimize the energy consumption of a building and the user comfort. These data are often strewn with rare or anomalous events, considered as anomalies (or outliers), that must be detected and ultimately corrected in order to improve the data quality. However, many approaches are used or might be used (for the most recent ones) to achieve this purpose. This paper proposes a classification methodology of anomaly detection techniques applied to building measurements. This classification methodology uses a well-suited anomaly typology and measurement typology in order to provide, in the future, a classification of the most adapted anomaly detection techniques for different types of building measurements, anomalies and needs.