Abstract: As the underlying bearing pipe of user service, it’s needed to analyze the network performance related monitoring data of packet optical network in real time and carry out targeted optimization operations to ensure the service performance. In view of the above control demands, based on the research of existing data analysis algorithms, this paper presented a LSTM+ VAE based packet optical network monitoring data analysis and optimization framework, including data training and visualization, characteristics analysis and optimization target calculation.At last, by collecting packet optical network delay data as experimental data, the effectiveness of the proposed method is verified. This research can also provide theoretical and experimental guidance for applyingArtificial Intelligence into the field of optical network management and control such as monitoring and optimization.

Key words: LongShortTermMemory(LSTM), optical networkmanagementand control,  , auto-encoder