Probabilistic Event-Driven Heuristic Fault Localization using Incremental Bayesian Suspected Degree
Cheng Zhang, Jianxin Liao, Xiaomin Zhu · 2008
Most fault localization techniques are based on time windows. The size of time windows impacts on the accuracy of the algorithms greatly. This paper took weighted bipartite graph as fault propagation model and proposed a heuristic fault localization algorithm based on incremental Bayesian suspected degree (IBSD) to eliminate the above shortcomings. IBSD sequentially analyzes the incoming symptoms in an event-driven way and incrementally computes the Bayesian suspected degree and determines the most probable fault set for the current observed symptoms. Simulation results show that the algorithm has high fault detection rate as well as low false positive rate and has a good performance even in the presence of unobserved alarms. The algorithm which has a polynomial computational complexity can be applied to large scale communication network.