Reinforcement Learning in Traffic Prediction of Core Optical Networks using Learning Automata

The development of technology related to optical core networks provides the appropriate infrastructure for the reliable flow of information, that is constantly increasing with the increase of interconnected users and devices and the evolution of the applications used. Network engineering following traffic prediction is one of the most important factors determining the effectiveness of core networks. Because of its importance, the research community has dealt extensively with traffic forecasting by proposing prediction mechanisms that are mainly based on time series and machine learning models. The objective of traffic forecasting mechanisms is to achieve the maximum possible accuracy while maintaining a low complexity in their operation. In this paper, we propose a novel traffic prediction mechanism that is based on reinforcement learning using Learning Automata. The evaluation of the proposed mechanism is carried out using traffic matrices derived from real measurements in a core network. Our performance simulation results show the effectiveness of the proposed mechanism, which achieves satisfactory forecast accuracy with low operational complexity.