Department of Systems and Computer Networks, Wrocław University of Science and Technology, Wrocław, Poland
The increasing demands placed on Elastic Optical Networks (EONs) underscore the importance of efficiently managing Routing and Spectrum Allocation (RSA). This paper advocates for the utilization of advanced machine learning techniques, specifically XGBoost models and stacking. By conducting a thorough analysis, it showcases the effectiveness of ensemble methods, particularly highlighting the superiority of stacking with linear and ridge regression as meta-learners, in accurately predicting network performance metrics. Additionally, the study delves into the delicate balance between computational efficiency and model accuracy, emphasizing the strategic application of ensemble learning for optimizing network performance in EONs.