Network Traffic Prediction Model Based on Convolutional Neural Networks‐Long Short‐Term Memory and iTransformer

ABSTRACT Accurately predicting network traffic is crucial for dynamically deploying computing resources in network data centers and reducing carbon emissions. In this paper, a hybrid prediction model Convolutional Neural Networks‐Long Short‐Term and iTransformer (CNN‐LSTM‐iTransformer) based on CNN‐LSTM and iTransformer is proposed. CNN—LSTM is used to capture local features and long—term dependencies, while iTransformer is employed for feature relevance learning and prediction. In addition, the Huber Loss function is used to further improve the model prediction accuracy. In the experiment, the dataset was provided by Ant Financial Group, and the experimental results show that CNN—LSTM—iTransformer significantly reduces MAE to 0.112, MSE to 0.0212, MAPE to 0.123, and RWMAPE which represents prediction risk to 0.122, so the hybrid model CNN‐LSTM‐iTransformer achieves not only a higher prediction accuracy but also a lower prediction risk.