A Novel Approach for Network Event Correlation Based on Set Covering

This paper proposes a novel network event correlation approach based on set covering. Firstly, we present a triple-layer belief network FPM by considering alarm loss and spurious faults into the bipartite graph FPM which used by IHU algorithm, then present the measurement of fault hypothesis. On basic of this model, we present a recursive algorithm for creating fault hypotheses. Compared with the IHU-based approach, the RHC algorithm can get all hypotheses which satisfy the relevance and non-redundant indexes during the hypotheses creating procedure. Simulations show the RHC algorithm is more accuracy than the IHU algorithm, especially in the circumstances with alarm losses and spurious alarms.