This study presents a machine learning framework for predicting pedestrian accident severity using Amman, Jordan's first complete 10-year traffic dataset (2014–2023). Addressing the critical class imbalance where minor injuries predominate (85%), causing standard models to poorly detect severe cases (< 25% recall), we implement cost-sensitive algorithms and specialized undersampling techniques, such as XGBoost with Balancing the Loss Function (XGBLF) and Random Data Partitioning with Voting Rule (RDPVR), which enhanced learning from underrepresented Major/Fatal cases while maintaining data authenticity. Through mixed-type correlation analysis and statistical testing, vehicle speed, road illumination, vehicle type, driver age, and road conditions emerged as the most significant predictive factors. RDPVR achieved a 63% true positive rate for Major/Fatal injuries, a 2.78-fold improvement over standard classifiers, and XGBLF achieved 95%, but this achievement was on account of the accuracy of the minor cases. Comprehensive interpretability analysis (SHAP, LIME, and Permutation Importance) revealed that heavy vehicles, poor lighting, and high-speed driving strongly predict Major/Fatal outcomes. Notably, the analysis demonstrates Jordan's improved safety trajectory, with 2020–2023 showing reduced severe accidents compared to 2014–2016, indicating measurable policy impact. This study delivers the first interpretable, context-sensitive AI framework for Amman/Jordan pedestrian safety, translating technical insights into actionable recommendations for targeted interventions, urban planning, and data-driven enforcement strategies to reduce pedestrian injury severity in high-risk zones.

Roundabouts are a prevalent type of intersection known for their potential to enhance traffic flow. Ensuring their effective design is crucial for optimizing traffic performance. This study focuses on evaluating the capacity of roundabouts, essential for both planning new installations and assessing existing ones. Field data from thirteen roundabouts in Hungary were analyzed to estimate critical gap and follow-up headway values for each entry. Employing Raff's graphical method, critical gap values were determined, while follow-up headway was calculated by averaging the time taken for two waiting vehicles to accept the same gap over eight instances. The critical gaps and follow-up headway values for all forty-one entries ranged between 2.41–3.46 s and 1.8–2.4 s, respectively. Subsequently, the Highway Capacity Manual (HCM) roundabout capacity equation was calibrated using these gap acceptance parameters. The proposed model yielded higher entry capacity (1,672 PCU/h) compared to the HCM model. Validation against actual field entry capacity values demonstrated a strong correlation (R2 = 0.94), affirming the model's accuracy. Comparisons with international models, such as HCM 2016, Brilon-Wu, and Brilon-Bondzio, revealed the superiority of the proposed model in terms of entry capacity (1,672 PCU/h versus 1,380 PCU/h, 1,241 PCU/h and, 1,218 PCU/h respectively).

This study investigates lateral lane-keeping behavior among human-driven passenger vehicles on Jordanian multi-lane roads. Using overhead video footage collected at five sites, 500 vehicles traveling alone in the leftmost lane under free-flow, daylight conditions were manually annotated for centerline deviation. The lateral position was analyzed using descriptive statistics, temporal trends, and spectral frequency analysis. Results show that 61% to 83% of vehicles remained within a ±0.5 m “safe zone” from the lane center. No vehicle exceeded the ±1.75 m legal lane boundary, and wheel position plots confirmed consistent lateral margins. Sites 1 and 2 exhibited a slight rightward bias, while Sites 3 through 5 showed a leftward tendency, especially Site 3, which had the highest variability (std dev = 0.43 m). Spectral analysis revealed consistent low-frequency oscillations (~0.01-0.02 Hz), indicating slow, smooth steering adjustments with no erratic corrections. The study confirms that under ideal conditions, drivers maintain stable lateral control within 3.5 m lane widths. These results provide valuable reference data for autonomous vehicle calibration, infrastructure planning, and future research into lane-keeping behavior under variable traffic and environmental conditions.

This study presents a comprehensive multidimensional analysis of pedestrian accidents in Amman, Jordan, from 2014 to 2023. By integrating spatial, temporal, and statistical techniques, the research identifies critical risk patterns to inform evidence-based safety interventions. Characterizing a decade-long database of 14,821 cases, the study utilizes radar graphs, Kernel Density Estimation (KDE), and DBSCAN cluster analysis to delineate high-risk zones and temporal peaks. Temporal findings indicate that Thursdays recorded the highest accident frequency (2382 cases), with peak occurrences between 17:00 and 23:00. Spatial clustering identified five significant high-risk zones, with Central Amman emerging as the primary critical area. The study’s novelty lies in being the first in the Jordanian context to bridge accident frequency with severity mechanisms by integrating advanced spatial clustering and KDE with a robust Ordered Logit Model. Severity analysis reveals that while 59.34% of incidents resulted in minimal injuries, fatalities accounted for 5.02%. The model demonstrates that injury outcomes are systematically associated with traffic dynamics and behavior rather than environmental factors. Speed-related driver error was identified as the strongest predictor of severe outcomes (OR = 81.3). Significant dependencies were confirmed between vehicle category and road type (χ² = 2182.20, p < 0.001), lighting and road surface (χ² = 76.21, p < 0.001), and vehicle type and lighting (χ² = 148.52, p < 0.001). The study proposes a multi-layered framework combining site-specific nodal improvements with corridor-level strategies to enhance urban safety in Amman City.