A machine learning framework for enhancing QoS in SD-WAN

Software defined networking has developed as potential paradigm that separates the control plane from the data plane, providing more adaptable and programmable network management. However, ensuring standard quality of service for various networking applications remains a significant challenge. Network latency, packet loss and ineffective path selection frequently affects network performance. This work introduces an advanced approach that combines traditional graph theory algorith ms with Artificial Intelligence (AI) algorithm to determine the optimal path that satisfy traffic requirements in terms of laten cy or bandwidth. The proposed AI model dynamically predicts the path expected to provide maximum bandwidth or predicts the path expected to provide minimum latency, then the selected path is established by applying the corresponding rules produced by the controller application. This work is implemented and tested using Python and Mininet environments. The results show that the proposed method achieved optimal ability to predict the best path for traffic flow. The accuracy of proposed model reached 96% in predicting the path with lowest latency and reached 94% in predicting the path with maximum bandwidth.