User behavior detection framework based on NBP for energy efficiency

In this paper, a user behavior detection framework based on non-parametric belief propagation (NBP) is proposed to discover the anomalous events from the dataset of contextual user behaviors, as in the case of anomalous events that can make a surge of energy flow and mess the energy management schedule, such as bursty occupancy, or unusual equipment usage. Firstly, the hybrid informationʹs fusion from multiple channels, which collects generalized information of contextual environment, is resolved by message discretization and update based on importance sampling, particle filters and non-parametric representation of multi-layer NBP. Then, the message passing process of belief propagation assists to identify the time span of anomalous events, and construct the contextual environment. Furthermore, we leverage the belief estimations of NBP inference to approximate the marginal probability of contextual anomalous events, and the Monte Carlo methods draw the samples of event states recursively to simulate the uncertainty of anomalies. Finally, the efficiency of NBP user behavior detection framework is validated by KDD2006 Calit2 Building Date Set.