Optical transmission system link error code prediction method and device based on machine learning

The invention relates to the technical field of optical transmission systems, and provides an optical transmission system link error code prediction method and device based on machine learning. Wherein important features which can cause link error codes are selected through the Gini importance of the random forest; performing modeling and machine learning on error code prediction based on maximum likelihood logistic regression by using the selected important features to train a model; and calculating the error probability of a future time period by using a model trained by machine learning and taking the important features of the selected link error codes as input. According to the invention, the analysis accuracy and robustness of the whole system are comprehensively improved. 本发明涉及光传输系统技术领域，提供了一种基于机器学习的光传输系统链路误码预测方法和装置。其中通过随机森林的基尼重要性选择会造成链路误码的重要特征；利用所选的重要特征基于最大似然的逻辑回归对误码预测进行建模和机器学习训练模型；使用机器学习训练好的模型以所选择链路误码的重要特征作为输入计算未来时段的误码概率。本发明综合提高了整个系统的分析准确性和鲁棒性。