Network Fault Detection Based on Bayesian Network
Yan Pu-liu · Wuhan University Journal · 2004
An intelligent fault detection approach based on Bayesian network is introduced. Individual MIB variable is modeled as a finite mixture model. Residual generated based on the model parameters is used to characterize the behavior of MIB variable. Then residuals of multiple MIB variables are combined in the probabilistic framework of a Bayesian network to compute the probability of fault, including unknown or unpredictable faults. Experiments are given to verify the validity of the approach.