Determinants influencing risks in e-bike cyclists under mix traffic condition: a partially constrained random parameters approach using experimental study data.

A spatially adaptive empirical Bayes framework with dynamic dispersion parameters for enhanced crash frequency prediction across rural highway networks.

Identifying environmental factors related to motorcyclist crash rates: variable selection using spatial Random Forest with network distance and barriers.

Spatial-temporal gated transformer network for freeway secondary crash prediction considering the impact of class imbalance.

Traffic crashes, especially rear-end collisions, fixed-object crashes, and rollovers, are common and severe, highlighting the need for comprehensive classification and management. However, research on classifying and managing imbalanced traffic crash data is limited. Therefore, this study proposes an analytic framework for imbalanced traffic crash type classification and management using a hybrid tabular transformer approach. The central idea is to analyze data using feature selection, multi-crash type classification, single-crash type clustering, and targeted safety recommendations. First, the multinomial logit model is used for feature selection, removing features with low correlation. Second, the Feature Tokenizer Transformer (FT-Transformer) with the Synthetic Minority Over-Sampling Technique (SMOTE) at varying ratios to perform multiclass crash classification is used. Third, mini-batch K-means clusters crash types based on key features are used. Finally, targeted safety recommendations for each cluster are developed. This study used 5,515 real-world traffic crash records from Anhui Province, China. The results show that: (1) the FT-Transformer model, with SMOTE at a 1:2:4 ratio, outperformed other machine learning models; and (2) rear-end, fixed-object, and rollovers were clustered into three, three, and five categories, respectively. Safety recommendations focus on traffic management (e.g., real time traffic updates), driver behavior (e.g., driving education), and road infrastructure (e.g., reinforced road markings).

Motorcycles are a common mode of transport in Ghana, such that in some regions, they account for 90% of all registered vehicles. Their popularity and associated crash vulnerabilities have become a significant public health issue. However, there is limited knowledge of the long-term epidemiological profile of fatal crashes in Ghana. This study examines the epidemiology of fatal motorcycle crashes over a 23-year period. We analysed 23 years of national data on motorcycle crashes. The data is collated by the Building and Road Research Institute from police-reported crash investigation files. Data were summarised for the temporal, environmental, and rider-related features of fatal crashes. A mixed-effects Poisson regression with robust variance estimation was employed to identify factors influencing fatal motorcycle crashes. Over the 23 years, 40,322 motorcycle crashes occurred, and 22.85% were fatal. Fatal crashes have increased, with an average annual percent change increase of 4.71% and 60% occurring from 2016 to 2022. The average age of the riders was 31.64 years, and 98.90% of the fatal crash riders were males. A greater proportion of fatality ratios were recorded in the northern parts Ghana (Upper West = 36.80% and Upper East = 36.60%), and during the last three months of the year (28.97%). The day of the week, time of day, weather, traffic control, road type, rider age and sex, and collision type were significant factors associated with fatal motorcycle crashes. The findings emphasise a significant public health burden from motorcycles, particularly over the past 10 years. Targeted interventions (e.g., speed-control measures) are needed to address this crisis, particularly as the country prepares to commercialise its use for passenger transport.

Falls due to disturbances are a common cause of serious cycling injuries, yet evaluation approaches to systematically evaluate interventions aimed at improving balance recovery are lacking. Current ex-post evaluations are hindered by sparse crash data, and existing ex-ante approaches often lack generalizability or rely on surrogate measures that are not validated against fall risk. This study introduces the Maximum Allowable Handlebar Disturbance (MAHD), a novel performance indicator that quantifies the largest handlebar disturbance a cyclist can recover from without falling. The MAHD captures the cyclist's resilience to disturbances and provides a direct, interpretable measure of intervention effectiveness. We propose two methods for determining MAHD: (1) controlled treadmill experiments with induced handlebar disturbances and safe fall conditions and (2) simulations using bicycle dynamics and cyclist control models. Together, these methods allow quantitative ex-ante evaluation and systematic comparison of interventions targeting cyclist control, bicycle design, and infrastructure features such as curbs and road shoulders. With further validation, the MAHD offers practical value for researchers, engineers, and policymakers seeking to design safer bicycles, training programs, and road environments and improve evidencebased resource allocation. In the future, this could reduce fall-related cycling injuries.

Rural freight mobility and logistics face persistent challenges, including inadequate road infrastructure, high transportation costs, safety risks, tolls at link access points, and dispersed demand. Traditional inventory routing models often fail to address these complexities, especially in rural contexts where alternative routing options and integrated in-haul/back-haul operations are essential for improving efficiency and reducing empty miles. This study proposes a bi-objective mathematical model for the inventory routing problem in rural logistics, incorporating multiple routing attributes (transportation costs, risks, link-access tolls, and distances) and inventory dynamics (integrated in-haul and back-haul visits). The model aims to minimize total logistics costs and accident risk while balancing operational expenses and safety considerations. Risk estimation is derived from crash data along rural road links connecting distribution nodes. A real-world case study involving Walmart distribution centers in Macclenny, Baker County, Florida, and several rural Supercenters is conducted to validate the model. A modified Non-Dominated Sorting Genetic Algorithm II (NSGA-II) is developed and compared with CPLEX for solution efficiency across small and large-scale problem instances. Results indicate that the proposed approach outperforms classical methods, improves routing decisions in rural logistics systems, and achieves cost savings of up to 17% for the evaluated objectives, emphasizing the importance of using multi-attribute, multi-route network structures in rural logistics optimization.

Evaluating roundabout safety in low- and middle-income countries is challenging due to unreliable crash data. This study addresses the issue by analyzing safety perception data from 1,530 questionnaire respondents across two Indian cities. A novel framework was developed to compare advanced ordinal ensemble models (Ordered XGBoost, LightGBM, Random Forest) against conventional ordinal regression (Logit, Probit). The ensemble models proved vastly superior, achieving Quadratic Weighted Kappa (QWK) scores exceeding 0.94, while conventional methods scored below 0.53. The top-performing Ordered XGBoost model (QWK=0.97) was interpreted using the state-of-the-art explainable artificial intelligence (XAI) technique SHAP (SHapley Additive exPlanations). SHAP analysis quantified the influence of key factors on perceived risk, identifying personal attributes (occupation, accident/near-accident experience) and infrastructure deficits (inadequate lighting, missing navigational aids) as primary drivers. The findings offer SHAP-quantified insights for deploying targeted, evidence-based safety interventions, providing a blueprint for improving perceived safety in complex traffic environments where traditional analysis is infeasible.

Thailand’s Western New Year and Songkran festivals witness a surge in traffic crashes due to increased travel volume. This study explores risk factors influencing crash injury severity during these holidays (2017–2019). Crash data is analyzed using both the Random Parameters Ordered Logit Model with Means Heterogeneity (RPOLHM) and machine learning techniques (Multilayer Perceptron (MLP), Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), Random Forest). Crash severity is categorized as property damage only (PDO), minor injury, and severe/fatal injury. The results reveal that factors like motorcycle or pedestrian involvement, adverse weather, speeding, drunk driving, fatigue, nighttime conditions, improper overtaking, and urban location all significantly increase the risk of severe/fatal crashes. Notably, the XGBoost model outperforms both RPOLHM and other machine learning methods, achieving a validation accuracy of 83.8%. While machine learning approaches demonstrate superior predictive capability, RPOLHM provides interpretable coefficient estimates and marginal effects essential for understanding causal mechanisms. This complementarity suggests that concurrent application of both paradigms offers comprehensive insights: machine learning for prediction-oriented objectives and econometric models for policy formulation. These findings provide valuable guidance for policymakers, highway engineers, and researchers to develop targeted road safety interventions during these high-risk periods.

The effect of license plate number-based vehicle restrictions on crash frequency.

Profiling crash-associated factors and injury risk patterns among lost-in-thought (daydreaming) drivers: a combined cluster-sequence analysis approach.

Pedestrian safety remains a pressing challenge in low- and middle-income countries (LMICs), where global predictive models often misrepresent local realities. This study tests the hypothesis that global predictive models, such as the International Road Assessment Programme (iRAP), overestimate countermeasure effectiveness in LMICs because key contextual factors are omitted. The two-phase research combined a PRISMA-based systematic literature review (SLR) with a quantitative iRAP performance gap analysis of the countermeasures implemented in the candidate studies of the SLR. The review systematically evaluated the effectiveness of pedestrian safety countermeasures, with an emphasis on their application in LMIC contexts. Following PRISMA 2020 guidelines, 14 longitudinal before–after studies were selected from 1911 records and screened with EPPI-Reviewer 4 software. The analysis identified 33 contextual factors shaping countermeasure performance across both high- and low-income settings; of these, 23 were specific to LMICs, and 13 are not accounted for in the iRAP model. The findings show that iRAP systematically overestimates countermeasure effectiveness in LMICs due to weak enforcement, poor maintenance, and informal road use. Transverse rumble strips were the only intervention consistently effective across diverse LMIC settings. A novel performance gap analysis of five LMIC case studies revealed an average discrepancy of 30.9% (SD = 29.7%) between predicted and observed outcomes. A risk of bias assessment showed that most LMIC studies were of moderate to serious risk, reflecting systemic data limitations and a frequent reliance on proxy outcomes. These findings highlight the urgent need for recalibrated, context-sensitive frameworks that incorporate enforcement, maintenance, and socio-economic variables. Policy implications include prioritising affordable and scalable countermeasures, pairing infrastructure with enforcement and education, and strengthening crash data systems to support more realistic, evidence-based road safety planning.

Ensuring pedestrian safety, especially for children near urban schools in developing countries, has always been critical in transportation and urban planning. Most prior research has employed area-based approaches with global statistical methods, which overlook spatial heterogeneity. This study explored how the built environment near schools affects pedestrian crashes, using both area-based and segment-based approaches. Eight models—global and local—were developed to analyze factors influencing pedestrian crashes near schools. Global models included Poisson regression (PR1) and negative binomial regression (NB2) for area-based analysis and binary logistic regression (LR1 and LR2) for segment-based analysis. Local models included geographically weighted Poisson (GP1 and GP2) and logistic (GL1 and GL2) regression for area-based and segment-based analyses, respectively. These models were applied using 5 years of crash data across two strategies: (1) school-age and (2) all-age. High-crash locations were identified using kernel density estimation based on significant variables. The results revealed that local area-based models (GP1 and GP2) have better accuracy than their corresponding global models. These local models showed that some variables (such as transit stop’ density and land use entropy) had location-specific effects. In global models, the all-age area-based model (NB2) demonstrated that some variables positively associate (arterial roads, transit stops’ density, average betweenness, and land use entropy) with crash counts during school start and end times. The segment-based models (LR1 and LR2) showed that some variables (arterial roads, average betweenness, one-way streets, and commercial land use) increase crash likelihood, while some others (residential land use, medians and pedestrian overpasses) reduce it.

Signalized intersections are frequently installed in developing countries to facilitate efficient traffic flow and seldom to increase traffic safety. As a result, fatal collisions still occur at intersections with signals. The purpose of this study is to gain a better understanding of signalized intersection safety by identifying and segmenting traffic and geometric risk factors associated with fatal crashes. For this purpose, a thorough road inventory survey—primary crash data—was used to analyze crashes at 67 signalized intersections in Hyderabad, an Indian metropolitan city. This paper proposes a multi-perspective model application and segmentation strategy that classifies a group of important crash factors determining crash fatality at urban signalized intersections by combining machine learning, data mining, and statistical modeling results. The proposed segmentation divided the crash parameters into three distinct categories: very high, high, and moderate risk factors. The key findings show that major road width, lack of right-turn protection, and absence of all-red time are the most influential factors contributing to fatal crashes at signalized intersections. Based on the findings, several policy recommendations were proposed. The segmentation of signalized intersection features would provide useful insights into the level of their influence and the impact of signalized intersection design on safety in developing countries. The study’s findings and proposed policy insights may assist transportation officials in developing, prioritizing, and implementing specialized safety countermeasures for signalized intersections.

Road crashes remain a major global concern, underscoring the importance of accurately identifying crash hotspots to support proactive safety management. The objective of this study is to examine the use of large-scale trajectories for crash hotspot identification. Five years of crash data and trajectory records totaling 36,531,563 observations from ten cities in China were analyzed. Four kinematic surrogate safety measures (SSMs), acceleration, deceleration, yaw rate, and jerk, were extracted from trajectories to generate trajectory-based hotspots using Kernel Density Estimation (KDE). The spatial consistency between trajectory-based hotspots and crash-derived hotspots was systematically evaluated to assess the effectiveness of different SSMs. The results show that trajectory-based hotspots substantially overlap with crash hotspots, demonstrating the feasibility of using trajectory-derived SSMs for hotspot identification. Among the evaluated measures, jerk consistently achieved the highest prediction accuracy and precision across cities. The contribution of this study is that it develops and empirically validates a trajectory-based crash hotspot identification framework and provides a systematic multi-city comparison of kinematic SSMs for hotspot identification. The findings provide practical implications for supporting proactive safety screening and informed decision-making in data-driven traffic safety management.

This study proposes a novel transfer learning framework, TabNet-Dynamic Adversarial Adaptation Networks (TabNet-DAAN), to address the challenge of inadequate and imbalanced traffic crash data in certain regions. By integrating TabNet’s robust feature extraction with the dynamic adversarial framework of DAAN, this approach enhances transfer learning effectiveness through automatic alignment of data distributions across different domains. Moreover, it integrates a Class-balanced Loss (CBLoss) algorithm to mitigate the adverse effects of data imbalance and further enhance the model’s robustness. The methodology was empirically validated using two-vehicle crash data (2018–2022) from Birmingham (7,432 crashes, source domain) and Leeds (3,275 crashes, target domain) in the UK. Results from ablation studies confirm the effectiveness of the CBLoss algorithm and TabNet module. Moreover, TabNet-DAAN demonstrates superior performance compared to other transfer and non-transfer learning algorithms in predicting injury severity of target domain data, exhibiting excellent generalization capabilities, particularly when using a 7:1 source-to-target data ratio in training. Additionally, feature importance analysis shows that TabNet-DAAN adaptively refines its feature selection, effectively identifying key predictors across domains. These findings highlight the model’s potential to improve predictions of injury severity in data-scarce regions, presenting profound implications for targeted traffic safety interventions.

Objectives We examined at-fault injury crashes of four passenger car populations: Hybrid Electric Vehicles (HEVs), Plug-in Hybrid Electric Vehicles (PHEVs), Battery Electric Vehicles (BEVs) and traditional internal combustion engine vehicles (ICEVs). For these populations, crash rates were calculated in relation to both registration years and mileage. Finally, controlled crash rate ratios were calculated to compare the crash risk between electric vehicles (EVs) and ICEVs. Methods Studied car populations were identified and their vehicle information for the period of 2019–2023, including the mileage (76 billion kilometers for all cars during the study period), was drawn from the national Vehicular and Driver Data Register. In addition, cars in the study populations were identified from the motor liability insurance (MLI) database and the crash data for them was retrieved (11,388 motor vehicle occupant injury crashes in total). Crash rates and crash rate ratios were calculated to evaluate the crash risk of EVs. Negative binomial regression was used to model crash involvement rate ratios both per registration year and per mileage for EVs, controlling the age and gender of the vehicle owner and vehicle size. Results Only battery electric vehicles showed significantly different crash rates than ICEVs per mileage, although the result was weakly significant −15% [−28%; 0%]. There were no significant differences in crash rates per registration years. In addition, there were only a few significant differences in crash circumstances between EVs and ICEVs. On average, the motor vehicle occupant injury crash rate of ICEVs was 151 crashes per billion kilometers and 2.37 crashes per thousand registration years. Conclusions Our results indicate that, when measured by motor vehicle occupant injury crash rate, passenger cars—regardless of powertrain—have not become safer in Finland compared to the situation ten years ago. However, the current crash rate of BEVs is lower than that of ICEVs. Previous studies suggest that some of the differences in crash rate may be explained by varying usage conditions, which our findings support. Part of the difference may be explained by differences in driver populations, which should be investigated further.

This study introduces a data-driven framework that integrates kinetic energy principles into proactive intersection safety management. Building on the Federal Highway Administration’s (FHWA) Safe System for Intersection (SSI) model, this research proposes the Kinetic Velocity Index (KVI), a model calibrated using categorized crash data to better represent the physical dynamics of intersection crashes. Unlike the SSI model, the KVI estimates the probability of fatal and serious injuries (FSIs) at the crash level, simplifying application while preserving alignment with safe system principles. The study used over 900,000 two-vehicle intersection crashes from the years 2013 to 2024. The model was developed using data from Georgia and validated with an external dataset from Massachusetts. The KVI maintained strong predictive performance (R 2 = 0.991 in Georgia; R 2 = 0.922 in Massachusetts), demonstrating its generalizability. Importantly, the KVI enables proactive safety screening by estimating severity risk even at locations with no prior FSIs, offering a more reliable and risk-oriented alternative to models that rely solely on past crash outcomes, which can be subject to random variation and may not reflect underlying crash risk. It also supports integration with historical data in weighted risk models. Because the KVI is easy to compute, interpretable, and built on physics-based principles, it provides transportation professionals with a more precise and scalable tool for identifying high-risk locations and evaluating the safety implications of design alternatives. Furthermore, the KVI framework is easily adaptable and can be locally recalibrated to reflect regional crash patterns, making it applicable across diverse roadway contexts.

Objective Blind spot monitoring systems help drivers avoid collisions with another vehicle when changing lanes. The systems provide visual and/or haptic warnings; active systems additionally avoid collision by applying brakes or steering the vehicle. This study aimed to estimate real-world effectiveness of these technologies as installed in the Australasian light vehicle fleet in Australasian road and driving conditions. Methods Police-recorded crash data were studied for the years 2019–2023 from the Australian states Victoria, Queensland, South Australia, New South Wales and Western Australia and also New Zealand. Using quasi-induced exposure analysis, rates of lane change crash involvements were studied for vehicles manufactured from 2018, classified by whether they were fitted with blind spot monitoring systems or not. Results We found a statistically significant 15% reduction (95% CI 26%-3%) in lane change crashes for vehicles equipped with a blind spot monitoring system. There was indicative evidence that the system was more effective for male than female drivers. For crashes involving injury, the associated reduction was estimated to be larger: a 24% reduction (95% CI 38%-6%). Active systems were rare in our data, but the analysis suggested they may be even more effective than the warning systems. Conclusions Despite the relatively low prevalence of lane change crashes generally, the estimated 15% reduction is significant. In the Australasian crash data analyzed, older drivers had an elevated rate of lane change crashes and may consequently gain greater benefit from blind spot monitoring systems.