Scholarlink

This paper constructs a prediction model based on multimodal spatio-temporal Bayesian neural network. The model integrates traffic flow, image and meteorological data, uses graph convolution to model spatial relationships, combines variational inference to achieve uncertainty quantification, and is built on the basis of GCN and LSTM for structural optimization. The experimental results show that the proposed model outperforms the traditional LSTM, Transformer and non-Bayesian multimodal models in terms of RMSE, MAE and R2, with the lowest RMSE of 12.03, the highest R2 of 0.941, and the coverage of 0.98 in the confidence interval prediction, which demonstrates good generalization and credibility.

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 Flow Prediction Based on Multimodal Spatio-Temporal Bayesian Neural Network

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zhangkang Tan,
  • Yongchao Shi,
  • Yuchen Zhang

摘要

This paper constructs a prediction model based on multimodal spatio-temporal Bayesian neural network. The model integrates traffic flow, image and meteorological data, uses graph convolution to model spatial relationships, combines variational inference to achieve uncertainty quantification, and is built on the basis of GCN and LSTM for structural optimization. The experimental results show that the proposed model outperforms the traditional LSTM, Transformer and non-Bayesian multimodal models in terms of RMSE, MAE and R2, with the lowest RMSE of 12.03, the highest R2 of 0.941, and the coverage of 0.98 in the confidence interval prediction, which demonstrates good generalization and credibility.