Network traffic forecasting plays a crucial role in intelligent network operations, but existing techniques often perform poorly when faced with limited data. Additionally, multi-task learning methods struggle with task imbalance and negative transfer, especially when modeling various service types. To overcome these challenges, we propose Sim-MSTNet, a multi-task spatiotemporal network traffic forecasting model based on the sim2real approach. Our method leverages a simulator to generate synthetic data, effectively addressing the issue of poor generalization caused by data scarcity. By employing a domain randomization technique, we reduce the distributional gap between synthetic and real data through bi-level optimization of both sample weighting and model training. Moreover, Sim-MSTNet incorporates attention-based mechanisms to selectively share knowledge between tasks and applies dynamic loss weighting to balance task objectives. Extensive experiments on two open-source datasets show that Sim-MSTNet consistently outperforms state-of-the-art baselines, achieving enhanced accuracy and generalization.
@article{arxiv.2601.21384,
title = {Sim-MSTNet: sim2real based Multi-task SpatioTemporal Network Traffic Forecasting},
author = {Hui Ma and Qingzhong Li and Jin Wang and Jie Wu and Shaoyu Dou and Li Feng and Xinjun Pei},
journal= {arXiv preprint arXiv:2601.21384},
year = {2026}
}