A bandwidth allocation scheme based on GRU traffic prediction in passive optical networks

With the advancement of information technology, network slicing technology has emerged as a viable solution for ensuring Quality of Service (QoS) in optical access networks. Current research is increasingly focusing on the integration of optical access networks with network slicing technologies. This paper proposes a bandwidth allocation scheme based on traffic prediction, specifically designed for resource management in optical network slicing scenarios. The scheme employs a Gated Recurrent Unit (GRU) neural network model to forecast future traffic, and combines bandwidth and latency factors to allocate bandwidth to each slice based on predicted values, thereby meeting the QoS requirements of various services. Simulation results indicate that, compared to the baseline algorithm, the proposed scheme achieved a 35.42% increase in Explaining Variance Score (EVS) and a 38.16% improvement in R2 score for factory slicing prediction. Similarly, for data center slicing prediction, the EVS score increased by 32.29% and the R2 score improved by 41.96%. In terms of performance metrics such as latency, packet loss rate, and throughput, the proposed algorithm outperforms both traditional prediction algorithms and the baseline algorithm.