A tutorial on the EM algorithm for Bayesian networks: Application to self-diagnosis of GPON-FTTH networks

摘要: Network behavior modelling is a central issue for model-based approaches of self-diagnosis telecommunication networks. There are two methods to build such models. The model can be built from expert knowledge acquired network standards and/or the learnt data generated by components mining algorithms. In recent work, we proposed architecture and fault propagation GPON-FTTH (Gigabit Passive Optical Network-Fiber To Home) access network. This based on Bayesian which encodes knowledge. includes dependencies that encode conditional probability distributions strength those dependencies. this paper consider problem automatically tuning above mentioned distributions. parameter estimation under missing conditions solve with Expectation Maximization (EM) algorithm. Conditional tremendous amount alarms an operating during months in 2015. Self-diagnosis carried out analyze root cause alarms. performance diagnosis evaluated respect system deterministic decision rules currently used Internet Access Provider diagnose problems.