Research on the fusion application of LSTM and transformer in network traffic prediction

In order to improve the accuracy of network traffic prediction, it is proposed to integrate LSTM and Transformer model to improve the accuracy of prediction. By combining LSTM and self-attention mechanisms, LSTM provides the ability to retain and transition sequence information, while self-attention mechanisms ensure that the model can focus on key features in the sequence regardless of distance. This combination makes the model more flexible and expressive when dealing with long sequences. The experimental results show that the r2 score of the fusion model is increased by 2.9% and 5.3%, and the Mape is increased by 1.55% and 1.94%, respectively, compared with the single LSTM model and Transformer model. The validity and rationality of the model are verified.