Learning partial correlation graph for multivariate sensor data and detecting sensor communities in smart buildings

Abstract

The storage and processing of massive time series data collected from smart buildings consume considerable computational resources. However, major information redundancy can be found in the smart building data. This paper proposed a partial correlation graph based approach to map the dependencies among sensors and detect the sensor communities in which the sensors are strongly “net” correlated. Specifically, the sparse partial correlation estimation method is used to learn the partial correlation graph. The Louvain algorithm is used to detect the communities of sensors by optimising the graph modularity. The case study demonstrates that the proposed method can identify spare sensors in the detected sensor communities and thus enhance the computational feasibility of smart building applications.