Scholarlink

Accurate traffic speed forecasting is a key component of intelligent transportation systems, supporting congestion management and travel time optimization in modern cities. This study investigates urban traffic as a multivariate time-series problem and conducts a benchmark evaluation across multiple real-world datasets. To address the lack of regional data diversity, we extend the benchmark with a new dataset collected from Ho Chi Minh City, reflecting complex dynamics in tropical urban environments. The proposed evaluation framework systematically analyzes temporal patterns, spatial correlations, and generalization behaviors across city-level networks. Experimental results confirm the importance of integrating spatial–temporal structures for robust forecasting and reveal practical challenges when transferring models between heterogeneous cities. The study contributes a unified benchmark setting and an open HCM traffic dataset (Ho Chi Minh City) for advancing research on urban traffic time-series forecasting.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Traffic Speed Forecasting in Urban Cities: A Time-Series Benchmark Evaluation and HCMC-Traffic Dataset Extension

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Thuan Q. Nguyen,
  • Hung P. Nguyen,
  • Nguyen Thu Nguyet Minh

摘要

Accurate traffic speed forecasting is a key component of intelligent transportation systems, supporting congestion management and travel time optimization in modern cities. This study investigates urban traffic as a multivariate time-series problem and conducts a benchmark evaluation across multiple real-world datasets. To address the lack of regional data diversity, we extend the benchmark with a new dataset collected from Ho Chi Minh City, reflecting complex dynamics in tropical urban environments. The proposed evaluation framework systematically analyzes temporal patterns, spatial correlations, and generalization behaviors across city-level networks. Experimental results confirm the importance of integrating spatial–temporal structures for robust forecasting and reveal practical challenges when transferring models between heterogeneous cities. The study contributes a unified benchmark setting and an open HCM traffic dataset (Ho Chi Minh City) for advancing research on urban traffic time-series forecasting.