Failure Prediction Based on LSTM and SVM under SDON Architecture

Artificial Intelligence in data center optical network is very popular these years. In this paper, we proposed an optical network failure prediction scheme based on long short-term memory (LSTM) and support vector machine (SVM) under Software Defined Optical Network (SDON) architecture. LSTM is used to predict parameters of the optical network in the future, and then SVM is used to judge whether the network will fail. The prediction accuracy can reach 0.9063 which means this method is efficient. When the failure is predicted to occur, the controller issues instructions to avoid the failure.