Interpretable machine learning for evaluating risk factors of freeway crash severity.

Machine learning (ML) models are widely employed for crash severity modelling, yet their interpretability remains underexplored. Interpretation is crucial for comprehending ML results and aiding informed decision-making. This study aims to implement an interpretable ML to visualize the impacts of factors on crash severity using 5 years of freeways data from Iran. Methods including classification and regression trees (CART), K-nearest neighbours (KNNs), random forest (RF), artificial neural network (ANN) and support vector machines (SVM) were applied, with RF demonstrating superior accuracy, recall, F1-score and ROC. The accumulated local effects (ALE) were utilized for interpretation. Findings suggest that light traffic conditions (volume/capacity<0.5) with critical values around 0.05 or 0.38, and higher proportion of large trucks and buses, particularly at 10% and 4%, are associated with severe crashes. Additionally, speeds exceeding 90 km/h, drivers younger than 30 years, rollover crashes, collisions with fixed objects and barriers, nighttime driving and driver fatigue elevate the likelihood of severe crashes.

Analyzing the transition from two-vehicle collisions to chain reaction crashes: A hybrid approach using random parameters logit model, interpretable machine learning, and clustering.

Chain reaction crashes (CRC) begin with a two-vehicle collision and rapidly intensify as more vehicles get directly involved. CRCs result in more extensive damage compared to two-vehicle crashes and understanding the progression of a two-vehicle collision into a CRC can unveil preventive strategies that have received less attention. In this study, to align with recent research direction and overcome the limitations of econometric and machine learning (ML) modelling, a hybrid approach is adopted. Moreover, to tackle the existing challenges in crash analysis, addressing unobserved heterogeneity in ML, and exploring random parameter effects and interactions more precisely, a new approach is proposed. To achieve this, a hybrid random parameter logit model and interpretable ML, joint with prior latent class clustering is implemented. Notably, this is the first attempt at using a clustering with hybrid modeling. The significant risk factors, their critical values, distinct effects, and interactions are interpreted using both marginal effects and the SHAP (SHapley Additive exPlanations) method across clusters. This study utilizes crash, traffic, and geometric data from eleven suburban freeways in Iran collected over a 5-year period. The overall results indicate an increased risk of CRC in congested traffic, higher traffic variation, and on horizontal curves combined with longitudinal slopes. Some parameters exhibit distinct or fluctuating effects, which are discussed across different conditions or considering interactions. For instance, during nighttime, heightened congestion on 2-lane freeways, increased traffic variation in less congested conditions, and adverse weather combined with horizontal curves and slopes pose risks. During daytime, increased traffic variation within highly congested sections, higher proportion of heavy vehicle traffic in moderately congested sections, and two lanes in each direction coupled with curves, elevate the levels of risk. The results of this study provide a better understanding of risk factors impact across different conditions, which are usable for policy makers.

Comparative analysis of pedestrian crash severity at United Kingdom rural road intersections and Non-Intersections using latent class clustering and ordered probit model.

Pedestrian safety is a critical issue in the United Kingdom (UK) as pedestrians are the most vulnerable road users. Despite numerous studies on pedestrian-vehicle crashes globally, limited research has been conducted to explore the factors contributing to such incidents in the UK, especially on rural roads. Therefore, this study aimed to investigate the severity of pedestrian injuries sustained on rural roads in the UK, including crashes at intersections and non-intersections. We utilized the STATS19 dataset, which provided comprehensive road safety data from 2015 to 2019. To overcome the challenges posed by heterogeneity in the data, we employed a Latent Class Analysis to identify homogeneous clusters of crashes. Additionally, we utilized the Ordered Probit model to identify contributing factors within each cluster. Our findings revealed that various factors had distinct effects on the severity of pedestrian injuries at intersections and non-intersections. Several parameters like the pedestrian location in footway and one-way roads are only statistically significant in the intersection section. Certain factors such as the day of the week, the pedestrian's location in a refuge, and minor roads (class B roads) were found to be significant only in the non-intersection section.Parameters includingpedestrians aged over 65 years and under 15 years, drivers under 25 years, male drivers and pedestrians, darkness, heavy vehicles, speed limits exceeding 96 km/h (60 mph), major roads (class A roads), and single carriageway roadsare significant in both sections. The study proposes various measures to mitigate the severity of pedestrian-vehicle crashes, such as improving lighting conditions, enhancing pedestrian infrastructure, reducing speed limits in crash-prone areas, and promoting education and awareness among pedestrians and drivers. The findings and suggested measures could help policymakers and practitioners develop effective strategies and interventions to reduce the severity of these incidents and enhance pedestrian safety.

Willingness and intention to run a red light among motorcyclists.

INTRODUCTION: An important issue affecting the safety of riders is running a red light. Many factors can affect this risky behavior including demographic, safety, and meteorological factors. METHOD: Using two models, the Prototype Willingness Model (PWM) and the Theory of Planned Behavior (TPB), this study examines this behavior. In this study, 853 participants completed an online questionnaire. RESULTS: The results indicated that older people and those who were married ran the red light less frequently compared with other riders. Additionally, people who rode motorcycles for more hours in a day had greater intentions and willingness to run the red light. People who had prior risky experiences while riding in the past reported running red lights more frequently. When comparing the two models, the PWM predicted a greater level of variance in the red-light running behavior when compared with the TPB. Among the components of the TPB model, the attitude factor was the strongest predictor. Prototype similarity was the strongest predictor of red-light running among the components of the PWM. Finally, it was observed that using a road safety educational (RSE) solution can effectively reduce the rate of red light traffic running among motorcyclists.

Importance - Performance Analysis (IPA) of metro service attributes during the COVID-19 pandemic.

Background: The study of user satisfaction with public transport services is of key interest for enhancing public transport ridership. However, the elevated risks of virus transmission on public transport associated with COVID-19 have resulted in a shift towards private vehicles. This issue is of concern to policymakers due to reduced levels of public transport use and increased private vehicle trips. Thus, it is relevant to re-evaluate attributes impacting user satisfaction during the COVID-19 pandemic to understand how this compares to related research undertaken before the pandemic. Methodology: Through a survey of 507 Tehran Metro users, the structural patterns among 33 metro service attributes were determined using Exploratory Factor Analysis (EFA). A service quality model of the Tehran Metro service (including cleanliness, ventilation/temperature, congestion, operation hours, ticket, information, reliability, and safety) was then designed using Confirmatory Factor Analysis (CFA). Attributes and latent constructs were classified using Importance - Performance Analysis (IPA). Findings: Attributes associated with general congestion in the metro, the performance of the metro during peak hours, cleanliness, and temperature/ventilation were found to be among the most influential attributes during the pandemic. Conclusion: Before the COVID-19 pandemic, time and cost effectiveness were key attributes affecting the choice of travel mode. Due to the outbreak of the virus, safety and health have been identified as the main attributes affecting the choice of travel mode. In this research, with a case study of the Tehran metro, it was found that there is a shift in attributes affecting service quality towards cleanliness, congestion and temperature/ventilation. As a result, a range of policies is needed to deal with the effect of COVID-19 on metro services. These include: increasing the number of metro services, reducing the time interval between metro services, improving air conditioning, and regular cleaning of vehicles. Policymakers can use the results of this study to improve user satisfaction with metro services and increase ridership levels during the pandemic.

Has COVID-19 changed our loyalty towards public transport? Understanding the moderating role of the pandemic in the relationship between service quality, customer satisfaction and loyalty.

COVID-19 has been associated with a decline in public transport ridership in many cities. However, the impact of the pandemic on passenger perceptions of public transport, particularly loyalty, has remained largely unexplored. Using a case study of Tehran's bus system, this paper aims to assess the moderating effect of COVID-19 on passenger loyalty. A cross-sectional survey of bus users was conducted in two similar periods, before and during COVID-19. The data was analysed using structural equation modelling, with the effect of observed heterogeneity evaluated using multiple indicators multiple causes models and multi-group analysis. The results indicate that during COVID-19, perceptions of service quality have improved, particularly in relation to comfort (including crowding), safety, reliability and information. While this has led to increased customer satisfaction, loyalty to the system has decreased. An increase in the perceived attractiveness of using private cars was found to be the primary factor for reducing passenger loyalty with the perceived monetary and psychological costs of using cars being less effective in encouraging people to use the bus. Taking into account the significant utility of private vehicles in urban transportation during COVID-19, the present study identifies feasible areas for improving the performance of bus services to increase satisfaction with this form of public transport among car owners.

Using ordered and unordered logistic regressions to investigate risk factors associated with pedestrian crash injury severity in Victoria, Australia.

INTRODUCTION: The safety of pedestrians is a major concern in Victoria, Australia. Despite the considerable number of pedestrian fatalities and injuries in traffic crashes, a limited number of studies focused on pedestrian crash severity in Victoria. METHODS: This study investigates and identifies the influential factors determining the severity of pedestrian injuries in traffic crashes in Victoria by using crash data from 2010 to 2019. An unordered multinomial logit model and an ordered logit model are developed for this purpose. RESULTS: The results indicate that pedestrian crashes on weekends, in the period of 10 a.m. to 10 p.m., on dark streets, at intersections, in areas with a speed limit above 50 km/h, and on medians or footpaths are associated with a higher probability of severe and fatal injuries. Male pedestrians, children, and older adults (>59) were more likely to sustain a higher level of injury in crashes. Concerning the driver characteristics, no significant relationship was found between pedestrian injury severity and driver gender and license status, but older drivers were more likely to cause severe and fatal injuries. Pedestrian collisions with motorcycles, heavy vehicles, light commercial vehicles, bus/minibus/coach, and trams increase the probability of more severe injuries compared to cars. Moreover, older vehicles are associated with a higher probability of severe pedestrian injuries. Comparison of the model results illustrated that the MNL model was slightly better fitted on the data than the ordered logit model, but the conclusions inferred from these two models were generally similar. PRACTICAL APPLICATION: To reduce the injuries of pedestrian crashes, we recommend improving lighting conditions and sidewalk design, implementing speed reduction strategies at high pedestrian activity areas, introducing more pedestrian crossings at midblock, installing warning signs to drivers, and discouraging the use of vehicles that are more than 20 years old.

Using latent class clustering and binary logistic regression to model Australian cyclist injury severity in motor vehicle-bicycle crashes.

INTRODUCTION: In recent years, Australia is seeing an increase in the total number of cyclists. However, the rise of serious injuries and fatalities to cyclists has been a major concern. Understanding the factors affecting the fatalities and injuries of bicyclists in crashes with motor vehicles is important to develop effective policy measures aimed at improving the safety of bicyclists. This study aims to identify the factors affecting motor vehicle-bicycle (MVB) crashes in Victoria, Australia and introducing effective countermeasures for the identified risk factors. METHOD: A data set of 14,759 MVB crash records from Victoria, Australia between 2006 and 2019 was analyzed using the binary logit model and latent class clustering. RESULTS: It was observed that the factors that increase the risk of fatalities and serious injuries of bicyclists (FSI) in all clusters are: elderly bicyclist, not using a helmet, and darkness condition. Likewise, in areas with no traffic control, clear weather, and dry surface condition (cluster 1), high speed limits increase the risk of FSI, but the occurrence of MVB crashes in cross intersection and T-intersection has been significantly associated with a reduction in the risk of FSI. In areas with traffic control and unfavorable weather conditions (cluster 2), wet road surface increases the risk of FSI, but the areas with give way sign and pedestrian crossing signs reduce the risk of FSI. Practical Applications: Recommendations to reduce the risk of fatalities or serious injury to bicyclists are: improvement of road lighting and more exposure of bicyclists using reflective clothing and reflectors, separation of the bicycle and vehicle path in mid blocks especially in high-speed areas, using a more stable bicycle for the older people, monitoring helmet use, improving autonomous emergency braking, and using bicyclist detection technology for vehicles.

Effects of COVID-19 on rail passengers' crowding perceptions.

Understanding changes in travel behavior during the spread of pandemic diseases such as COVID-19 is important to develop a resilient transportation system. Since one of the most important ways to prevent the spread of this virus is to keep a safe distance from other people (i.e. social distancing), it has implications for the operations of public transportation as compared to other modes of transportation due to the confinement of a large number of passengers in enclosed space. This study investigated the effect of the spread of COVID-19 on crowding perception and crowding disutility in metro rail system of Tehran. Two surveys were conducted before and during the COVID-19. The stated preference data were analyzed by mixed logit models with the lognormal distribution. Results revealed that the value of crowding increased during the pandemic. Tracking the changes of crowding perception caused by COVID-19 shows that low comfort scores were observed at crowding levels where seats were taken, and the density of standees was high (i.e. not possible to maintain social distancing). During the outbreak of COVID-19, crowding has more disutility for rail passengers and the value of having a seat while traveling increased. Understanding passengers' perceptions of crowding as examined in this study will assist transport operators, and planners maintain the critical functionality of public transport systems and manage risks of mass transportation during the pandemic and beyond.

Modelling bus-pedestrian crash severity in the state of Victoria, Australia.

Understanding the relationship between bus-pedestrian crash severity and factors contributing to such crashes is important. However, there exists a dearth of research on the factors affecting bus-pedestrian crash severity. This study aims to fulfil this gap by investigating the factors affecting the severity of pedestrian injuries. A data set of bus-pedestrian crashes in the State of Victoria, Australia was analysed over the period of 2006 - 2019. Through the results of association rule discovery method, the factors that increase the risk of pedestrian fatality are darkness, pedestrian walking on carriageway with traffic, intersections, high speed zone, old pedestrian, young bus driver and weekend holidays. Furthermore, co-occurrence of factors that increase the risk of a pedestrian fatality were extracted. To reduce the injuries of bus-pedestrian crashes, we recommend improving the light conditions, reducing the jaywalking behaviour of pedestrians, implementing speed bumps in high speed zones and installing pedestrian detection systems on buses.13 years of bus-pedestrian crashes in Victoria, Australia was analyzed.Association rules discovery was used for modeling pedestrian fatality.Darkness, pedestrian movement, zone speed and age effect the rate of fatality.Pattern of pedestrian fatality in collision with bus was extracted.

Data mining approach to model bus crash severity in Australia.

INTRODUCTION: Buses are different vehicles in terms of dimensions, maneuverability, and driver's vision. Although bus traveling is a safe mode to travel, the number of annual bus crashes cannot be neglected. Moreover, limited studies have been conducted on the bus involved in fatal crashes. Therefore, identification of the contributing factors in the bus involved fatal crashes can reduce the risk of fatality. METHOD: Data set of bus involved crashes in the State of Victoria, Australia was analyzed over the period of 2006-2019. Clustering of crash data was accomplished by dividing them into homogeneous categories, and by implementing association rules discovery on the clusters, the factors affecting fatality in bus involved crashes were extracted. RESULTS: Clustering results show bus crashes with all vehicles except motor vehicles and weekend crashes have a high rate of fatality. According to the association rule discovery findings, the factors that increase the risk of bus crashes with non-motor vehicles are: old bus driver, collision with pedestrians at signalized intersections, and the presence of vulnerable road users. Likewise, factors that increase the risk of fatality in bus involved crashes on weekends are: darkness of roads in high-speed zones, pedestrian presence at highways, bus crashes with passenger car by a female bus driver, and the occurrence of multi-vehicle crashes in high-speed zones. Practical Applications: The study provides a sequential pattern of factors, named rules that lead to fatality in bus involved crashes. By eliminating or improving one or all of the factors involved in rules, fatal bus crashes may be prevented. The recommendations to reduce fatality in bus crashes are: observing safe distances with the buses, using road safety campaigns to reduce pedestrians' distracted behavior, improving the lighting conditions, implementing speed bumps and rumble strips in high-speed zones, installing pedestrian detection systems on buses and setting special bus lanes in crowded areas.

Investigating the interaction between pedestrian behaviors and crashes through validation of a pedestrian behavior questionnaire (PBQ).

Developing countries have a high frequency of traffic incidents involving pedestrians. Given the vulnerability of pedestrians, many of these incidents result in serious or fatal injuries. The present study aimed to validate a pedestrian behavior questionnaire in Iran to investigate Persian pedestrian behaviors and to understand the relationship of these behaviors with demographic and mobility variables. A total of 520 participants (292 males and 228 females) completed a survey containing behavioral items and demographic questions. A principal component analysis showed that the data best fit in four factors of transgressions (including violations and errors), lapses, aggressive behaviors, and positive behaviors. In this study, the association of behavioral dimensions with crash history as a driver, crash history as a pedestrian, the severity of an experienced pedestrian-related crash, and the participant's relative's crash history as a pedestrian were also investigated. The present study confirms that PBQ is a useful tool with adequate reliability for investigating Persian pedestrians' safety-related behaviors. These findings revealed the need for intervention programs and improving infrastructures through the evaluation of pedestrian behaviors, which may lead to decreasing pedestrian-related crash frequency.

Development of the pedestrian anger expression inventory.

OBJECTIVE: There has been extensive research into road user behavior, although how pedestrians express their anger has yet to be explored. This is important given the high vulnerability of pedestrians and the additional risks that angry people often undertake. The present study developed a tool to measure one's tendency to engage in aggressive acts while walking: The pedestrian anger expression inventory (PAX). METHODS: The survey consisted of 37 items adapted from the Driving Anger Expression Inventory as well as a subset of items from the Pedestrian Behavior Questionnaire. Ten items from the Trait Anger Scale (TAS) questionnaire were also used to assess the general anger tendencies of individuals. A total of 475 participants from Tehran provided complete responses to a questionnaire administered via paper and pencil. RESULTS: A Principal Component Analysis showed a 30-item, 3-factor model describing three ways of expressing anger: 1) Anger Expression-In (internalizing anger), 2) Anger Expression-Out (in the form of aggression), and 3) Adaptive/Constructive Expression (dealing with anger in a constructive way). Hierarchical linear regression showed that trait anger was a significant predictor of pedestrian anger expression, above other demographic variables (age and gender). However, age and gender remained significant predictors of pedestrian anger. As age increased, the tendency to become angry while walking decreased. Male pedestrians were also significantly more likely to express their anger aggressively than female pedestrians. CONCLUSIONS: In this study, we developed a questionnaire to measure anger expression in pedestrians, highlighting three broad ways pedestrians deal with their anger (internally, externally, or constructively). This questionnaire was used in Iran and therefore, further research is required to validate these tools among different samples and populations.

Observational-based study to explore pedestrian crossing behaviors at signalized and unsignalized crosswalks.

Despite many studies on exploring the behaviors of pedestrians crossing the road, there is a need for comprehensive studies that identify the factors that may influence pedestrians crossing behavior at signalized and unsignalized intersections. This study aims to comprehensively examine the influence of gender, age group, group-crossing, technological devices and carrying items on pedestrians crossing behaviors at signalized and unsignalized crosswalks simultaneously. Observational data of 552 pedestrians at two signalized and two unsignalized crosswalks in Tehran were collected. Temporal and spatial violations, conflict experience and collision avoidance, situational awareness, and pedestrians crossing speed were used as pedestrians crossing behaviors indicators. To model crossing behaviors, linear mixed models (LMMs) and Generalized linear mixed models (GLMMs) with fixed-effect approach were applied for the continuous outcome (pedestrians' crossing speed) and binary outcomes, respectively. Phi and Cramer's V coefficients were used to avoid multicollinearity. Results showed that traffic checks before and while crossing showed a high positive correlation with crossing at "don't walk" and flashing "don't walk" signals and conflict experience at signalized crosswalks. As compared to females, males started their crossing more on flashing "don't walk" signal and crossed the crosswalk faster. Older pedestrians exhibited more cautious behaviors at signalized intersections but less in unsignalized intersections. Alone pedestrians behaved more cautiously than groups and crossed the crosswalk more quickly. Using technological devices, regardless of their types, caused pedestrians to not exhibit one or more safe crossing indicators considered in this study. Pedestrians talking on their phones had the least cautious behaviors. Pedestrians listening to music mostly looked at the ground or straight direction rather than looking left-right for traffic. The findings from this study are a valuable resource to road authorities and policy makers to develop appropriate targeted strategies to prevent pedestrians' injuries and fatalities and improve crosswalks safety.

A multinomial logit model of motorcycle crash severity at Australian intersections.

INTRODUCTION: Motorcyclists are exposed to more fatalities and severe injuries per mile of travel as compared to other vehicle drivers. Moreover, crashes that take place at intersections are more likely to result in serious or fatal injuries as compared to those that occur at non-intersections. Therefore, the purpose of this study is to evaluate the contributing factors to motorcycle crash severity at intersections. METHOD: A data set of 7,714 motorcycle crashes at intersections in the State of Victoria, Australia was analyzed over the period of 2006-2018. The multinomial logit model was used for evaluating the motorcycle crashes. The severity of motorcycle crashes was divided into three categories: minor injury, serious injury and fatal injury. The risk factors consisted of four major categories: motorcyclist characteristics, environmental characteristics, intersection characteristics and crash characteristics. RESULTS: The results of the model demonstrated that certain factors increased the probability of fatal injuries. These factors were: motorcyclists aged over 59 years, weekend crashes, midnight/early morning crashes, morning rush hours crashes, multiple vehicles involved in the crash, t-intersections, crashes in towns, crashes in rural areas, stop or give-way intersections, roundabouts, and uncontrolled intersections. By contrast, factors such as female motorcyclists, snowy or stormy or foggy weather, rainy weather, evening rush hours crashes, and unpaved roads reduced the probability of fatal injuries. Practical Applications: The results from our study demonstrated that certain treatment measures for t-intersections may reduce the probability of fatal injuries. An effective way for improving the safety of stop or give-way intersections and uncontrolled intersections could be to convert them to all-way stop controls. Further, it is recommended to educate the older riders that with ageing, there are physiological changes that occur within the body which can increase both crash likelihood and injury severity.