Broadband Access Server Load Prediction Based on ICEEMDAN-NRBO-BiLSTM-Attention

The Broadband Access Server (BAS) is a critical network infrastructure that handles multiple tasks, including network access authentication, bandwidth control, and address allocation. The massive influx of wireless and IoT devices into the network has led to frequent fluctuations in the load on BAS. To address this, we propose a load prediction model for BAS based on ICEEMDAN-NRBO-BiLSTM-Attention. The ICEEMDAN decomposition is employed as a data preprocessing step before the BiLSTM-Attention model, decomposing the time series into multiple simpler sub-series, which are more easily learned and predicted by the BiLSTM-Attention network. The NRBO optimization algorithm is used to optimize the hyperparameters of the BiLSTM-Attention network, saving a significant amount of time and enhancing model performance. Validation on real datasets has demonstrated the model's high prediction accuracy.