Saved in:
Bibliographic Details
Main Authors: Ma, Hui, Li, Qingzhong, Wang, Jin, Wu, Jie, Dou, Shaoyu, Feng, Li, Pei, Xinjun
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.21384
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914289776852992
author Ma, Hui
Li, Qingzhong
Wang, Jin
Wu, Jie
Dou, Shaoyu
Feng, Li
Pei, Xinjun
author_facet Ma, Hui
Li, Qingzhong
Wang, Jin
Wu, Jie
Dou, Shaoyu
Feng, Li
Pei, Xinjun
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2601_21384
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Sim-MSTNet: sim2real based Multi-task SpatioTemporal Network Traffic Forecasting
Ma, Hui
Li, Qingzhong
Wang, Jin
Wu, Jie
Dou, Shaoyu
Feng, Li
Pei, Xinjun
Machine Learning
Artificial Intelligence
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.
title Sim-MSTNet: sim2real based Multi-task SpatioTemporal Network Traffic Forecasting
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2601.21384