Application of Probabilistic Modeling and Machine Learning to the Diagnosis of FTTH GPON Networks

This paper presents insights on the promises of probabilistic modeling and machine learning for fault diagnosis in optical access networks. A Bayesian inference engine, called Probabilistic tool for GPON-FTTH Access Network self-DiAgnosis (PANDA), is applied to fault diagnosis of Gigabit capable Passive Optical Networks (GPON). PANDA approach has been assessed on real diagnosis data, showing very satisfactory alignment with an operational rule-based expert system. Furthermore, it provides diagnosis conclusions for all tested cases, even if some monitoring data is missing or incomplete. Finally, an expectation maximization algorithm allows to finely tune the probabilistic model.