EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Short‐Term Traffic Volume Prediction for Expressways Based on the Bo‐CNN‐LSTM Combined Model

Xiaolei Li, Hongwen Wu, Qianghui Song, Yili Chen, Zexu Zhang

IET Intelligent Transport Systems · 2026-01 · 20 (1)

DOI: 10.1049/itr2.70285 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IET Intelligent Transport Systems, online erschienen am 28.07.2026. Verfasst von Xiaolei Li und 4 weiteren Autor:innen. Die Arbeit nennt 73 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate short‑term expressway traffic volume prediction is critical for enhancing traffic efficiency and enabling proactive management. However, existing methods often fail to fully consider the nonlinear influence of traffic composition and meteorological factors, and lack sufficient validation of model robustness against sensor noise. To address these limitations, a hybrid model integrating Bayesian optimisation (BO), convolutional neural networks (CNNs) and long short‑term memory (LSTM) is proposed, with the lagged PLV and meteorological features incorporated as explicit input features. Firstly, meteorological data are linearly interpolated, feature selection is performed using Pearson correlation and mutual information, a 7‑step cumulative lagged average PLV is constructed to capture delayed effects, and all data are normalised. Then, spatial features are extracted by CNN, temporal dependencies are captured by LSTM, and key hyperparameters are automatically tuned by BO. Finally, based on real expressway data, four evaluation metrics are employed, comparisons with seven baseline models are conducted, and ablation experiments together with noise robustness tests are carried out. The results demonstrate that after adding the lagged PLV, RMSE is reduced from 57.41 to 49.78 (a 13.5% decrease), and R 2 is increased from 0.6408 to 0.7300; the removal of BO causes an RMSE increase of 11.69. Under 10% Gaussian noise, RMSE increases by only 0.10 and R 2 decreases by only 0.0012, validating the strong robustness of the model. The proposed method offers a high‑accuracy, robust and easily deployable solution for short‑term traffic volume prediction without requiring graph‑based network structures.

Zitieren

APA

Li, X., Wu, H., Song, Q., Chen, Y., & Zhang, Z. (2026). Short‐Term Traffic Volume Prediction for Expressways Based on the Bo‐CNN‐LSTM Combined Model. IET Intelligent Transport Systems, 20(1). https://doi.org/10.1049/itr2.70285

Vancouver

Li X, Wu H, Song Q, Chen Y, Zhang Z. Short‐Term Traffic Volume Prediction for Expressways Based on the Bo‐CNN‐LSTM Combined Model. IET Intelligent Transport Systems. 2026;20(1). doi:10.1049/itr2.70285

BibTeX
@article{Li2026,
  author = {Li, Xiaolei and Wu, Hongwen and Song, Qianghui and Chen, Yili and Zhang, Zexu},
  title = {Short‐Term Traffic Volume Prediction for Expressways Based on the Bo‐CNN‐LSTM Combined Model},
  journal = {IET Intelligent Transport Systems},
  year = {2026},
  volume = {20},
  number = {1},
  doi = {10.1049/itr2.70285}
}
Kostenlos starten →