Traffic Flow Prediction in Data-Scarce Regions: A Transfer Learning Approach

Traffic flow prediction is a key component of intelligent transportation systems, particularly in data-scarce regions where traditional models relying on complete datasets often fail to provide accurate forecasts. These regions are characterized by limited sensor coverage and sparse data collection, pose significant challenges for existing prediction methods. To address this, we propose a novel transfer learning framework called transfer learning with deep knowledge distillation (TL-DKD), which combines graph neural network (GNN) with deep knowledge distillation to enable effective knowledge transfer from data-rich to data-scarce domains. Our contributions are three-fold: (1) We introduce, for the first time, a unique integration of deep knowledge distillation and transfer learning, enhancing feature adaptability across diverse traffic datasets while addressing data scarcity. (2) We design an encoder-decoder architecture where the encoder retains generalized spatiotemporal patterns from source domains, and the decoder fine-tunes predictions for target domains, ensuring minimal information loss during transfer. (3) Extensive experiments on five real-world datasets (METR-LA, PeMS-Bay, PeMS03/04/08) demonstrate the framework’s robustness. The TL-DKD model achieves significant improvements in prediction accuracy, especially in data-scarce scenarios. For example, the PEMSD4 dataset in multi-region experiments, it achieves a mean absolute error (MAE) of 20.08, a mean absolute percentage error (MAPE) of 13.59%, and a root mean squared error (RMSE) of 31.75 for 30-min forecasts. Additionally, noise-augmented experiments show improved adaptability under perturbed data conditions. These results highlight the framework’s practical impact, offering a scalable solution for accurate traffic predictions in resource-constrained environments.