RBFNN Based Demand Forecasting for Next Generation Ethernet Passive Optical Network

The rapid growth of internet users and bandwidth-intensive applications have led to increasing demand for efficient data transmission. Optical fiber transmission has become essential in network architecture and has adopted deep learning based intelligence to categorize complex internet traffic with a focus on bandwidth prediction. To improve the performance of the Next-Generation Ethernet Passive Optical Network, the proposed scheme uses a deep learning-based Radial Basis Function Neural Network (RBFNN) termed RBFNN-DBA model to track user's demand and predict their bandwidth needs before receiving a request from the optical network unit. By reducing the sole dependency on the traditional Request-Grant mechanism, the RBFNN-DBA model leads to assure a better quality of service metrics. The effectiveness of the RBFNN-DBA model is evaluated by comparing it to the existing long short-term memory model based on the grant-to-reporting ratio, end-to-end delay, and fairness. The results show that the proposed model outperformed all metrics.