Parameter Efficient Hybrid Transformer (PEHT) for Network Traffic Prediction via Dynamic Urban Congestion Integration
The Parameter-Efficient Hybrid Transformer (PEHT) is a network traffic prediction framework that integrates urban mobility and congestion data into a Transformer-based architecture. It isolates primary communication features from secondary mobility features, using Low-Rank Adaptation (LoRA) in the encoder to minimize trainable parameters, and fuses multimodal congestion data in the decoder. The model was evaluated using the Telecom Italia Milan dataset alongside synthetic congestion scenarios, outperforming baseline models in RMSE, MAE, and R-squared metrics.
Why it matters — It demonstrates that cellular network traffic forecasting can be made highly parameter-efficient without sacrificing accuracy by decoupling core communication data from dynamic urban mobility context.
Caveat: The framework's performance under real-world, non-synthetic congestion dynamics remains untested beyond the simulated scenarios and the single Milan dataset.