Abstract: Abstract: Traffic research is at the core of modern urban traffic management and sustainable development. Accurate traffic forecasting can not only improve the efficiency and safety of the traffic system, but also promote the sustainable development of society and economy. Although a large number of studies have been devoted to modeling spatiotemporal correlation, existing methods still have significant limitations: most models tend to collectively predict the traffic volume of all regions in all time periods, ignoring spatiotemporal heterogeneity, especially the impact of the traffic status of the current region on the future traffic status of related regions. To address this problem, a new method, Time-Interdependency-Aware Dynamic Bayesian Network for Traffic Prediction (TIDBG), is proposed. TIDBG uses a time-varying dynamic Bayesian network through a pre-training module to capture the complex temporal relationships in time series data due to simultaneous and lagged effects. To further improve the ability to capture spatiotemporal correlation, a spatiotemporal attention mechanism is introduced for in-depth analysis. Subsequently, a graph convolutional network (GCN) is used to model the spatiotemporal topological structure to generate more accurate traffic forecasts. Experimental results show that TIDBG performs well in two real traffic forecasting tasks. Especially in the 1-hour forecast, TIDBG improves the evaluation index MAE by 4% compared with the state-of-the-art methods.

Key words: traffic prediction, Bayesian network, graph convolutional network, attention mechanism, deep learning

摘要： 摘 要: 交通研究是现代城市交通管理与可持续发展的核心。精准的交通预测不仅能够提升交通系统的效率与安全性，还能促进社会和经济的可持续发展。尽管已有大量研究致力于建模时空相关性，但现有方法仍然存在显著的局限性：大多数模型倾向于集体预测所有区域在所有时间段的流量，忽略了时空异质性，特别是当前区域的交通状态对相关区域未来交通状态的影响。为了解决这一问题，提出了一种新方法，基于时间依赖建模的动态贝叶斯网络交通预测方法(TIDBG)。TIDBG通过预训练模块，利用时变动态贝叶斯网络捕捉时间序列数据中由于同时和滞后影响而产生的复杂时序关系。为进一步提升对时空相关性的捕捉能力，引入了时空注意力机制进行深入分析。随后，采用图卷积网络(GCN)对时空拓扑结构进行建模，以生成更准确的交通预测。实验结果表明，TIDBG在两个真实交通预测任务中均表现优异。尤其在1小时预测中，TIDBG相比最先进的方法，评价指标MAE提高了4%。

关键词: 交通预测, 贝叶斯网络, 图卷积网络, 注意力机制, 深度学习