Diagnosis, Optimization, and Verification of Localized Failures in Traffic Flow Prediction Models

DOI

10.2991/978-94-6239-648-7_76How to use a DOI?

Keywords

Traffic flow forecast; Stacking integrated learning; Data heterogeneity; Feature engineering; Abnormal value handling

Abstract

In the real traffic control scenario, the multi-intersection prediction model often appears “local failure” due to the distribution difference and concept drift between intersections, which leads to the decline of the overall prediction performance and affects the reliability of scheduling. Improving the robustness of the model under the conditions of distribution change and scene migration has become an important research direction of traffic time series modeling. This study proposes a data-driven framework of “diagnosis optimization verification” to be applied to the task of traffic flow prediction at multiple intersections. Aiming at the obvious local failure of stacking ensemble model (SEM), which is composed of linear regression and extreme gradient boosting (XGBoost), at junction 3, this paper analyzes the reasons for the decline of positioning performance through exploratory data analysis (EDA), including data distribution offset, extreme value interference and time series mode instability. In this study, targeted outlier truncation, multi-scale time series feature enhancement, box discretization and super parameter automatic optimization strategies were used to reduce the root mean square error (RMSE) of the model to 5.96 (43.49%) and improve the coefficient of determination (R²) to 0.73. This study provides a repeatable and interpretable system optimization path to deal with the phenomenon of “same model with different effects” in traffic forecasting, and has reference value for time series forecasting tasks with heterogeneous distribution of traffic, energy and so on.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yunjian Tang
PY  - 2026
DA  - 2026/04/24
TI  - Diagnosis, Optimization, and Verification of Localized Failures in Traffic Flow Prediction Models
BT  - Proceedings of the International Workshop on Advances in Deep Learning for Image Analysis and Computer Vision (IWADIC 2025)
PB  - Atlantis Press
SP  - 702
EP  - 712
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6239-648-7_76
DO  - 10.2991/978-94-6239-648-7_76
ID  - Tang2026
ER  -