One of the most notable advances in the telecommunications are the passive optical network (PON) protocols in the access segment. For example, the GPON protocol operates at 2.5 Gbps per optical port, or up to 10, 40, and now 50 Gbps, in both point-to-point and point-to-multipoint topologies, allowing Internet service users to enjoy higher bandwidths. The aforementioned protocols are defined in the recommendations of the International Telecommunication Union as G-PON under G.984, XG-PON (asymmetric) and XGS-PON (symmetric) under G.9807, or the most recent 50G-PON under G.9804. Current management interfaces for these protocols only allow for the identification of signal absence or degradation problems without further details and only at the time of the incident. This work proposes to predict the conditions that lead to signal degradation using artificial intelligence model to analyze the eye diagram under normal and different degradation conditions. This data is used to train neural networks that can estimate the conditions that lead to irregular operation of PON before they happens. Matlab and Python tools will be used to implement the neural networks.