Network Traffic Forecasting Model Based on Improved Fireworks Algorithm Optimized CNN-BiGRU

In response to the challenges encountered by existing models in network traffic prediction, particularly in terms of prediction accuracy and the difficulty of hyperparameter tuning, we propose a method that utilizes the Improved Fireworks Algorithm to optimize the hyperparameters of a combined Convolutional Neural Network and Bidirectional Gated Recurrent Units model, referred to as IFA-CNN-BiGRU. This model leverages CNN for feature extraction from data and BiGRU for capturing temporal dependencies, taking advantages of deep learning to better capture nonlinear relationships and long-term dependencies in time series data, thereby enhancing prediction accuracy. IFA algorithm realizes the automatic adjustment of model hyperparameters and enhances the generalization and adaptive ability of the model. Compared with manual hyperparameter tuning, the IFA-CNN-BiGRU model effectively optimizes the model structure. The results show that the RMSE of IFA-CNN-BiGRU is about 4.81% higher than that of FA-CNN-BiGRU and about 19.44% higher than that of CNN-BiGRU. This shows that IFA-CNN-BiGRU model has certain advantages in the task of network traffic prediction, and provides an effective solution to solve the limitations of existing models.