Research on Network Slicing Resource Prediction Method Based on Ring Federated Learning

This paper proposes a prediction method for network slicing resources based on ring federated learning. By obtaining the performance index data required for network slicing service in each VNFs sequence, and then training the network slicing resource prediction model according to the performance index data required for network slicing service in each VNF continuous time. Get a prediction of the resources required for network slicing. Among them, the network slicing resource prediction model is constructed by using the ring federated learning framework. The physical node where each VNF resides is taken as the client participating in federated learning, and the node resource prediction model is constructed for each client. Then, the performance indicator data required for the network slicing service in each VNF continuous time is input into the node resource prediction model of the physical node where the VNF is located, and the resource prediction results required at the next moment of network slicing are output, which improves the accuracy of network slicing resource prediction.