Cellular Traffic Prediction Based on Spatiotemporal Graph Feature Fusion

With the rapid development of 4G and 5G networks, accurate BS traffic prediction has become a critical factor for efficient cellular network management. Traditional traffic prediction models, such as Long Short-Term Memory (LSTM) and (Convolutional Neural Network) CNN+Transformer, either neglect spatial information or fail to efficiently integrate temporal and spatial features, limiting their prediction accuracy. To address this issue, we propose a Multi-Temporal Spatial Graph Convolution Network (MTSGCN) model. This model introduces a novel mechanism of multi-length historical sequences and combines a Gated Linear Unit (GLU)-based temporal module with a Graph Convolution Layer (GCL)-based spatial module, capturing both temporal and spatial dependencies effectively. Extensive experiments on real-world datasets demonstrate that MTSGCN achieves superior convergence and predictive accuracy compared to the existing models.