Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network
Accurately predicting heavy freight vehicle (HFV) traffic flow is essential for optimizing service levels and improving road safety management. This study compares the performance of three machine learning approaches—classical learning, reinforcement learning, and artificial neural networ...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11097318/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849391666908102656 |
|---|---|
| author | Alireza Gholami Seyedehsan Seyedabrishami |
| author_facet | Alireza Gholami Seyedehsan Seyedabrishami |
| author_sort | Alireza Gholami |
| collection | DOAJ |
| description | Accurately predicting heavy freight vehicle (HFV) traffic flow is essential for optimizing service levels and improving road safety management. This study compares the performance of three machine learning approaches—classical learning, reinforcement learning, and artificial neural networks—in predicting the traffic flow of trucks and tractor-trailers. The dataset consists of hourly traffic records spanning three years and seven months on rural road network segments in southern Iran. The modeling framework integrates temporal and spatial variables along with lagged traffic volumes from upstream counters. Model performance is assessed through an expanding window cross-validation approach, which defines separate training, validation, and testing datasets. The results indicate that while the seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) model performs well on the training data, it exhibits limitations in predicting test data. The extreme gradient boosting (XGBoost) model surpasses the time-series model in predictive accuracy, yielding average R-squared values of 84.7% and 85.8% on the test data for trucks and tractor-trailers, respectively. In comparison, the multi-layer perceptron (MLP) model achieves average R-squared values of 86.8% for trucks and 82.9% for tractor-trailers. In addition to R-squared, model performance has been further evaluated using RMSE and MAE metrics, offering a more nuanced and reliable understanding of prediction accuracy across different error dimensions. Analysis based on SHAP values and sensitivity assessment demonstrates that particular times of day and traffic counter classifications exert the greatest influence on HFV traffic flow predictions. Overall, the MLP model demonstrates superior predictive performance. |
| format | Article |
| id | doaj-art-05badb0b672a487a805461313b91bdf3 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-05badb0b672a487a805461313b91bdf32025-08-20T03:40:59ZengIEEEIEEE Access2169-35362025-01-011313579113580510.1109/ACCESS.2025.359320211097318Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road NetworkAlireza Gholami0https://orcid.org/0000-0002-3747-896XSeyedehsan Seyedabrishami1https://orcid.org/0000-0002-9301-5417Department of Civil and Environmental Engineering, Tarbiat Modares University, Tehran, IranDepartment of Civil and Environmental Engineering, Tarbiat Modares University, Tehran, IranAccurately predicting heavy freight vehicle (HFV) traffic flow is essential for optimizing service levels and improving road safety management. This study compares the performance of three machine learning approaches—classical learning, reinforcement learning, and artificial neural networks—in predicting the traffic flow of trucks and tractor-trailers. The dataset consists of hourly traffic records spanning three years and seven months on rural road network segments in southern Iran. The modeling framework integrates temporal and spatial variables along with lagged traffic volumes from upstream counters. Model performance is assessed through an expanding window cross-validation approach, which defines separate training, validation, and testing datasets. The results indicate that while the seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) model performs well on the training data, it exhibits limitations in predicting test data. The extreme gradient boosting (XGBoost) model surpasses the time-series model in predictive accuracy, yielding average R-squared values of 84.7% and 85.8% on the test data for trucks and tractor-trailers, respectively. In comparison, the multi-layer perceptron (MLP) model achieves average R-squared values of 86.8% for trucks and 82.9% for tractor-trailers. In addition to R-squared, model performance has been further evaluated using RMSE and MAE metrics, offering a more nuanced and reliable understanding of prediction accuracy across different error dimensions. Analysis based on SHAP values and sensitivity assessment demonstrates that particular times of day and traffic counter classifications exert the greatest influence on HFV traffic flow predictions. Overall, the MLP model demonstrates superior predictive performance.https://ieeexplore.ieee.org/document/11097318/Trucks and tractor-trailerstraffic flow predictionSARIMAXXGBoostMLP |
| spellingShingle | Alireza Gholami Seyedehsan Seyedabrishami Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network IEEE Access Trucks and tractor-trailers traffic flow prediction SARIMAX XGBoost MLP |
| title | Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network |
| title_full | Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network |
| title_fullStr | Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network |
| title_full_unstemmed | Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network |
| title_short | Leveraging Spatial and Temporal Data to Predict Heavy Freight Vehicle Traffic Flow on Rural Road Network |
| title_sort | leveraging spatial and temporal data to predict heavy freight vehicle traffic flow on rural road network |
| topic | Trucks and tractor-trailers traffic flow prediction SARIMAX XGBoost MLP |
| url | https://ieeexplore.ieee.org/document/11097318/ |
| work_keys_str_mv | AT alirezagholami leveragingspatialandtemporaldatatopredictheavyfreightvehicletrafficflowonruralroadnetwork AT seyedehsanseyedabrishami leveragingspatialandtemporaldatatopredictheavyfreightvehicletrafficflowonruralroadnetwork |