Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan
Abstract Reliable estimation of crash frequency at roundabouts is foundational to data-driven roadway safety analysis and design. However, conventional Safety Performance Function (SPF) development is typically constrained by the availability of long-term empirical crash data, limiting its applicabi...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Journal of Engineering and Applied Science |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s44147-025-00675-z |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849763215354888192 |
|---|---|
| author | Diana Al-Nabulsi Aya Hassouneh |
| author_facet | Diana Al-Nabulsi Aya Hassouneh |
| author_sort | Diana Al-Nabulsi |
| collection | DOAJ |
| description | Abstract Reliable estimation of crash frequency at roundabouts is foundational to data-driven roadway safety analysis and design. However, conventional Safety Performance Function (SPF) development is typically constrained by the availability of long-term empirical crash data, limiting its applicability in rapidly urbanizing or data-limited regions. This study introduces an advanced calibration framework for SPFs using machine learning techniques, demonstrated through a case study of 20 urban roundabouts in Amman, Jordan. A comprehensive dataset was assembled, incorporating annual average daily traffic (AADT) and a suite of geometric parameters including inscribed diameter, entry width, number of legs, circulating width, and splitter island radius. Comparative modeling was conducted using ordinary least squares regression and Random Forest Regressor algorithms. The linear regression model yielded an R 2 of 0.542 with a high sum of squared errors (SSE = 3750.38), underscoring its limited capacity to capture non-linear relationships. In contrast, the Random Forest model achieved an R 2 of 0.992, RMSE of 0.089, and MAE of 0.06, reflecting excellent model fit and predictive accuracy. A key innovation of this research lies in demonstrating that robust, data-driven SPF models can be pre-calibrated using high-resolution geometric and traffic inputs without the need for localized long-term crash records corresponding to the same features and region. This approach allows transportation engineers and policymakers to anticipate crash risks, evaluate design alternatives, and implement safety interventions in advance of incident accumulation. The findings offer a scalable, transferable framework for proactive roundabout safety planning, particularly in regions where traditional SPF development remains infeasible due to data limitations. |
| format | Article |
| id | doaj-art-db50d35d94e4485db786f2b656dc36bd |
| institution | DOAJ |
| issn | 1110-1903 2536-9512 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | Journal of Engineering and Applied Science |
| spelling | doaj-art-db50d35d94e4485db786f2b656dc36bd2025-08-20T03:05:29ZengSpringerOpenJournal of Engineering and Applied Science1110-19032536-95122025-07-0172111710.1186/s44147-025-00675-zDevelopment and calibration of roundabout safety performance functions using machine learning: a case study from Amman, JordanDiana Al-Nabulsi0Aya Hassouneh1Department of Civil and Construction Engineering, Western Michigan UniversityDepartment of Electrical and Computer Engineering, Western Michigan UniversityAbstract Reliable estimation of crash frequency at roundabouts is foundational to data-driven roadway safety analysis and design. However, conventional Safety Performance Function (SPF) development is typically constrained by the availability of long-term empirical crash data, limiting its applicability in rapidly urbanizing or data-limited regions. This study introduces an advanced calibration framework for SPFs using machine learning techniques, demonstrated through a case study of 20 urban roundabouts in Amman, Jordan. A comprehensive dataset was assembled, incorporating annual average daily traffic (AADT) and a suite of geometric parameters including inscribed diameter, entry width, number of legs, circulating width, and splitter island radius. Comparative modeling was conducted using ordinary least squares regression and Random Forest Regressor algorithms. The linear regression model yielded an R 2 of 0.542 with a high sum of squared errors (SSE = 3750.38), underscoring its limited capacity to capture non-linear relationships. In contrast, the Random Forest model achieved an R 2 of 0.992, RMSE of 0.089, and MAE of 0.06, reflecting excellent model fit and predictive accuracy. A key innovation of this research lies in demonstrating that robust, data-driven SPF models can be pre-calibrated using high-resolution geometric and traffic inputs without the need for localized long-term crash records corresponding to the same features and region. This approach allows transportation engineers and policymakers to anticipate crash risks, evaluate design alternatives, and implement safety interventions in advance of incident accumulation. The findings offer a scalable, transferable framework for proactive roundabout safety planning, particularly in regions where traditional SPF development remains infeasible due to data limitations.https://doi.org/10.1186/s44147-025-00675-zSafety performance functionsRandom Forest regressorRoundaboutGeometric featuresTraffic safety |
| spellingShingle | Diana Al-Nabulsi Aya Hassouneh Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan Journal of Engineering and Applied Science Safety performance functions Random Forest regressor Roundabout Geometric features Traffic safety |
| title | Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan |
| title_full | Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan |
| title_fullStr | Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan |
| title_full_unstemmed | Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan |
| title_short | Development and calibration of roundabout safety performance functions using machine learning: a case study from Amman, Jordan |
| title_sort | development and calibration of roundabout safety performance functions using machine learning a case study from amman jordan |
| topic | Safety performance functions Random Forest regressor Roundabout Geometric features Traffic safety |
| url | https://doi.org/10.1186/s44147-025-00675-z |
| work_keys_str_mv | AT dianaalnabulsi developmentandcalibrationofroundaboutsafetyperformancefunctionsusingmachinelearningacasestudyfromammanjordan AT ayahassouneh developmentandcalibrationofroundaboutsafetyperformancefunctionsusingmachinelearningacasestudyfromammanjordan |