Crash Severity Levels Research Articles

New insights into factors affecting the severity of autonomous vehicle crashes from two sources of AV incident records

Superior safety is the main banner value of promoting autonomous vehicle (AV) technology, but it is difficult to responsibly claim it. The potential for AVs to reduce crash and injury risks would be overshadowed by technological limitations, regardless of their ability to mitigate or eliminate human error. This study aims to identify the key factors affecting crash severity by analyzing real-world AV crash data from the U.S. between 2015 and 2022. We integrated two open data sources from the California DMV and NHTSA. Mixed multinomial logit models incorporating the interaction effects were estimated using the crash severity level, addressing the observed and unobserved heterogeneities. Our results show that Advanced Driver Assistance System (ADAS) engagement is associated with a higher likelihood of slight injury crashes, whereas Automated Driving System (ADS) engagements show the opposite effect. In addition, we found that various environmental and road factors, including lighting conditions, weather, road type, and road surface conditions, significantly affect the severity of AV crashes. For instance, daylight conditions contribute to a lower likelihood of slight-injury crashes. On the other hand, driving under unfavorable weather conditions (cloudy, foggy, rainy, or snowy), on the highway, and on non-dry road surfaces are associated with an increase in the likelihood of severe injury crashes. Furthermore, several significant interaction effects were revealed. First, the mitigating effect of ADS engagement on the likelihood of slight injury crashes is reduced by the rear-end collision type. Second, the likelihood of slight injury crashes increases when AV interacts with heavy trucks on highways. Third, the likelihood of severe injuries increases when AVs collide with Vulnerable Road Users (VRUs) on urban streets. Overall, this research is expected to provide policymakers and AV manufacturers with valuable insights into AV safety, stressing that addressing the identified factors will lead to improved AV design and control algorithms.

Enhancing work zone crash severity analysis: The role of synthetic minority oversampling technique in balancing minority categories

Road work zones are becoming increasingly common due to the aging infrastructure and the need for capacity enhancement. They present significant safety risks due to narrow lanes, uneven traffic flow, lower speed, and reduced visibility. It is particularly important to understand the role of human behavioral factors in WZ crash injury severity due to difficulty navigating such areas. Furthermore, the crash injury data available is mostly imbalanced, primarily due to the lower incidence of high-cost fatal and severe injuries, and can benefit from the use of emerging analysis techniques. This research study examines a unique dataset comprising 7,855 WZ crashes in Tennessee from 2018 to 2022 as a case study to provide useful insight into the behavioral factors associated with injury severity and how they change after adjusting for the underrepresented fatal and serious injuries within the dataset. The study applies frequentist methods and a machine learning technique enhanced with the Synthetic Minority Oversampling Technique (SMOTE), addressing the data imbalance (relatively fewer fatal and serious injuries) for useful inferences and predictions. The study results indicate that aggressive driving, overspeeding, and drunk driving significantly elevate injury severity. Additionally, after balancing the minority categories of crash injury severity levels, the importance of contributing factors changes. The study offers engineers and data analysts a framework for analyzing imbalanced data, a prevalent issue in crash injury severity analysis. By exploring key behavioral factors responsible for injury severity in WZ crashes, the study provides useful insight and valuable information to traffic safety engineers, transportation agencies, and policymakers to implement enhanced safety measures in WZ design and management, ultimately aiming to mitigate injury severity and to improve overall safety for road users.

Investigation of a surrogate measure-based safety index for predicting injury crashes at signalized intersections

Objectives The paper develops a machine learning-based safety index for classifying traffic conflicts that can be used to estimate the frequency of signalized intersection crashes, with a focus on the more severe ones that result in fatal and severe injury. The number of conflicts in different severity levels categorized by the safety index is used as an explanatory variable for developing statistical models for pro-actively estimating crashes. Methods Video-derived conflicts in different severity levels between left-turning vehicles and opposing through vehicles, a well-recognized severe injury crash typology at signalized intersections, were identified by jointly integrating the indicators of frequency and severity, using an autoencoder neural network integration method to develop anomaly scores. Regression models were then developed to relate crashes at the same intersections to the classified conflicts based on the value of their safety indexes. Cumulative Residual plots were investigated. Finally, equations defining the boundary between consecutive anomaly score levels were developed to facilitate application in practice. Results Regression models for total and fatal plus severe (FSI) crashes utilizing classified extreme conflicts based on anomaly scores were found to outperform the models using total conflicts. The improvement is more pronounced for FSI crashes. The results also suggest that the machine learning integration method can efficiently classify conflicts accurately according to crash severity levels since the higher anomaly score is associated with a higher crash severity level (i.e., FSI). Conclusions The proposed framework represents a methodological advancement in traffic conflict-based estimation of crashes using a machine learning model to classify conflicts by their anomaly scores. For jurisdictions without the resources to develop such a model to classify conflicts for their own datasets, the simple equations defining the boundary between consecutive anomaly score levels could be used as an approximation.

Classification of autonomous vehicle crash severity: Solving the problems of imbalanced datasets and small sample size

Only a few researchers have shown how environmental factors and road features relate to Autonomous Vehicle (AV) crash severity levels, and none have focused on the data limitation problems, such as small sample sizes, imbalanced datasets, and high dimensional features. To address these problems, we analyzed an AV crash dataset (2019 to 2021) from the California Department of Motor Vehicles (CA DMV), which included 266 collision reports (51 of those causing injuries). We included external environmental variables by collecting various points of interest (POIs) and roadway features from Open Street Map (OSM) and Data San Francisco (SF). Random Over-Sampling Examples (ROSE) and the Synthetic Minority Over-Sampling Technique (SMOTE) methods were used to balance the dataset and increase the sample size. These two balancing methods were used to expand the dataset and solve the small sample size problem simultaneously. Mutual information, random forest, and XGboost were utilized to address the high dimensional feature and the selection problem caused by including a variety of types of POIs as predictive variables. Because existing studies do not use consistent procedures, we compared the effectiveness of using the feature-selection preprocessing method as the first process to employing the data-balance technique as the first process.Our results showed that AV crash severity levels are related to vehicle manufacturers, vehicle damage level, collision type, vehicle movement, the parties involved in the crash, speed limit, and some types of POIs (areas near transportation, entertainment venues, public places, schools, and medical facilities). Both resampling methods and three data preprocessing methods improved model performance, and the model that used SMOTE and data-balancing first was the best. The results suggest that over-sampling and the feature selection method can improve model prediction performance and define new factors related to AV crash severity levels.

Crash severity prediction and interpretation for road determinants based on a hybrid method

In this study, 4 classification algorithms were employed to characterize the influence of road determinants on roadway crash severity with actual crash data. The crash data were obtained from crash records in Texas, USA, from January 2020 to April 2021. The prediction model of crash severity utilized 12 road-related features—including shoulder types, shoulder width, and curb types—as well as 10 other features—such as weather and illumination conditions—as input features. Three crash severity levels—“Minor Damage,” “Moderate Damage,” and “Severe Damage”—were used as output features. Decision tree, support vector machines, and multi-layer perceptron were employed to compare their prediction performance with the XGBoost model. The results show that the XGBoost model yields the best performance among the 4 algorithms. The overall accuracy, average precision, average recall, and average F1 score of the XGBoost model were 82.65%, 0.83, 0.82, and 0.82, respectively. Besides, SHapley Additive exPlanations (SHAP) and partial dependence plots were used to interpret the model results. Among the road-related features, the most influential one is the median width. Greater crash severity is related to paved right shoulder and curb. These findings are helpful for the design and planning of road safety.

Learning from Moped Crash Data: Identifying Risk Factors Contributing to the Severity of Injuries Sustained by Moped Riders

Mopeds are light vehicles that provide solutions for short-distance travel, traffic congestion, and environmental pollution issues. However, because of the limited physical protection they offer, moped riders are more likely to sustain severe injuries in a crash than vehicle passengers. This study aims to investigate the features and patterns associated with recent moped crashes and analyzed the severity of injuries sustained by riders and related risk factors in 1,657 moped crashes in Michigan during a five-year period (2017–2021). A multinomial logit model was built to examine the association between crash injury severity levels and 13 variables from three main categories: crash features; rider behaviors/cognitive status; and environmental conditions. The descriptive statistics showed that 79% of injuries sustained by moped riders in crashes were minor, and the injured riders tended to be younger, male, and not wearing helmets. The findings indicate a significant correlation between the severity of injuries sustained by riders and several factors: the type of crash (whether single, head-on, or angle collision); the age of the riders (with older riders being more susceptible), visibility conditions (especially during nighttime); and helmet usage. Notably, single and head-on crashes were the primary causes of injuries. These results highlight the importance of heightened vigilance among moped riders, particularly when navigating poorly maintained surfaces or riding under low-light conditions. It is imperative that riders consistently use helmets to minimize risks. Additionally, setting guidelines or restrictions on helmet use, rider age, riding hours, and permissible areas in which to ride could further contribute to improving overall safety for moped riders.

Investigating effective factors on rural crash severity at marginal areas around cities in Iran: a partial proportional odds modelling approach

Over the past decade, rural crashes have been responsible for an average of 65% of crash-induced casualties in Iran. Evidence from prior studies reveals that a significant number of these rural crashes occur at marginal areas around cities. Thus, Exclusive crash severity models should be developed to identify the factors associated with higher injury and fatality probabilities in these areas. In this study, a partial proportional odds (PPO) model was formulated using the rural crash data collected from roads leading to the city of Isfahan. The PPO model holds the ordinal nature of crash observations and allows for different influences of independent variables on various crash severity levels. Insights derived from the results reveal that factors such as vehicle traffic maintaining an average speed exceeding 95 km/h, the occurrence of multi-vehicle crashes, the incidence of overturn-type crashes, the at-fault vehicle being a truck/trailer and at-fault or not-at-fault vehicle being a motorcycle, increase the likelihood of more severe rural crashes. Conversely, a foreign vehicle being at-fault, and the driver of the at-fault vehicle aged between 30 and 40 years, tend to diminish the occurrence of severe crashes at marginal areas around cities.

"Road Crash Severity Level in Urban Areas. A Study on Traffic Crashes in the Romanian Cities "

Road traffic in urban areas is highly prone to the occurrence of road crashes, while crash severity level differs at the city level and in relation to the leading cause. In the case of urban sustainable environments, strategic planning and development measures have been implemented for safer mobility by effective traffic management and road user education. Still, in crowded cities, where mobility and transit traffic are major factors, road crashes are the antithesis of road safety and they occur in simple or complex circumstances mostly related to road users and road infrastructure. The aim of this study was to define the severity level of road traffic crashes in the cities of Romania during the reference period 2008 - 2019, before the COVID-19 pandemic, while highlighting the leading causes of road crashes in urban areas that determine their highest occurrence and severity level. Official crash data records provided by the general Directorate of the Romanian Police were employed. The general severity degree of road crashes was calculated, particularly in relation to the first five leading causes, and cities were ranked accordingly. Results are rendered at the local level, for all 319 cities and towns in Romania, by the five leading causes and by severity level, comparatively for the years 2008 and 2019. Synthetically, we pinpoint the Romanian cities that are most affected by this phenomenon. This could help local and national decision-makers direct their focus towards mitigating the negative effects of the frequent crash triggering factors, decrease road crash severity levels and create customized measures for better traffic management and increased road safety at the local level.

Analyzing the factors affecting the crash severity level at urban roundabouts in non-lane-based heterogeneous traffic

Pivotal crash factors are investigated, and crash-severity model for the safety assessment at roundabouts and its vicinity in non-lane based heterogenous traffic is developed. An ordered-probit model was developed using crash-data collected between 2015–2019 for 20 roundabouts in India. The analysis revealed critical influencing parameters for determining the severity-level of crash outcomes at roundabouts, namely, inscribed-circle diameter, height of central island, number of circulatory lanes, presence of splitter island and median, posted-speed limit, type of collision, type of violation behaviour, collision partner, the pattern of collision, presence of road lane-marking, presence of street-light and age of victims. To precisely quantify the impact of each significant factor, marginal effects analysis was also carried out. The results show that the probability of fatal-injuries increased by 14.28% due to angle-collision, 15% for hit-pedestrians, 20.6% due to the pattern of collision and 15.60% due to the collision-partner, Whereas the probability of occurrence of grievous injury was the highest for rear-end with 17%, followed by sideswipe collision with 16% respectively. This study’s findings can aid in developing effective remedies to reduce the crash severity for roundabouts road-users and updating the roundabout design standards, considering the safety perceptive.

Identification of determinant factors for crash severity levels occurred in Addis Ababa City, Ethiopia, from 2017 to 2020: using ordinal logistic regression model approach

BackgroundRoad traffic Injuries (RTI) are multifaceted occurrences determined by the combination of multiple factors. Also, severity levels of injuries from road traffic accidents are determined by the interaction of the composite factors. Even though most accidents are severe to fatal in developing countries, there is still a lack of extensive researches on crash severity levels and factors associated with these accidents. Hence, this study was intended to identify severity levels of road traffic injuries and determinant factors in Addis Ababa City, Ethiopia.MethodsThe study was conducted in Addis Ababa, the capital city of Ethiopia, using secondary data obtained from the Addis Ababa Police Commission office. The ordinal logistic regression model was used to investigate road traffic injury severity levels and factors worsening injury severity levels using the recorded dataset from October 2017 to July 2020.ResultsA total of 8458 car accidents were considered in the study, of which 15.1% were fatal, 46.7% severe, and 38.3% were slight injuries. The results of the ordinal logistic regression analysis estimation showed that being a commercial truck, college and above level educated driver, rollover crash, motorbike passengers, the crash day on Friday, and darkness were significantly associated factors with crash injury severity levels in the study area. On the contrary, driving experience (> 10 years), passenger of the vehicle, two-lane roads, and afternoon crashes were found to decrease the severity of road traffic injuries.ConclusionsRoad traffic injury reduction measures should include strict law enforcement in order to maintain road traffic rules especially among commercial truckers, motorcyclists, and government vehicle drivers. Also, it is better to train drivers to be more alert and conscious in their travels, especially on turning and handling their vehicles while driving.

Multivariate Approach to Peak-Period Models for Crash Types and Severities in Variable Speed Systems on Freeways

According to current research, safety performance functions (SPFs) based on traditional annual average daily traffic are less suitable for addressing crash risk in traffic-management strategies that typically operate for short time intervals (e.g., peak period). Variable speed limit (VSL)/variable advisory speed (VAS) is a traffic-management strategy which is deployed to achieve speed harmonization typically during the peak hours of weekdays. However, peak-period SPFs for VSL/VAS-implemented freeways considering different crash types and severity levels are still unexplored. Therefore, in this study the multivariate Poisson lognormal (MVPLN) method was applied to estimate the different crash-type and severity-level models. The results showed that by addressing correlations among different crash types and severity levels, MVPLN models outperformed individual univariate Poisson lognormal crash-type and -severity models. From the developed models, several traffic- and roadway-related attributes—such as volume, average speed, standard deviation of speed, segment type (merge, diverge, and weaving), higher number of lanes, and so forth—could be identified as the potential crash-risk factors. Also, the proposed models were successful in capturing the spatial effects of segments (differences between segments downstream and upstream of the location of interest) on crash frequency. Further, the implementation of VSL/VAS strategy was found to be effective in reducing rear-end crashes in the crash-type model as well as property-damage-only (PDO) and C (non-incapacitating) crash severity levels in the crash-severity model. Policymakers and practitioners could benefit from this study in incorporating safety aspects into developing VSL/VAS algorithms.

A novel one-vs-rest consensus learning method for crash severity prediction

Research in crash severity prediction is necessary to allow safety planners to take precautionary measures and enable first aiders to remain prepared for assisting the injured. Existing literature in the field of crash severity prediction is mostly focused on generating the attributes for predicting the severity. However, in reality, not all features are discriminating, and certain classes are challenging to detect even employing the entire feature set. Although to tackle these problems several techniques are developed in the Machine Learning (ML) literature. But their application to crash severity prediction and an optimal strategy for the best combination of features or classifiers for achieving high accuracy is a less studied area. To address these problems, this work first provides a comparison of widely used classifiers for predicting crash severity; and secondly, by combining class-wise majority voting with One-vs-Rest (OvR) approach, a novel classification framework named, OvR consensus learning (OvRCL) is proposed. The proposed method avail a feature selection technique, Mutual information (MI), to acquire the most relevant feature set regarding the output class (i.e. severity). Moreover, to differentiate each class in the multi-class data, OvRCL iteratively runs ML algorithms as binary classifiers in an ensemble framework to significantly ameliorate classification performance. In our experiments, we use four classifiers, namely, the k-Nearest Neighbors (k-NN), Support Vector Machines (SVM), Random Forest (RF), and Bagging classifier, to get the consensus. Analysis was done using a real crash dataset obtained from an open data source of Leeds city council. A four-year crash data (2015–2018) is used for training and the OvRCL is tested on the 2019 data. Moreover, to validate the performance of OvRCL, this study also utilizes two more datasets with high-class imbalance. In contrast to conventional ML algorithms, our experiments depictthat the OvRCL is a potent method for forecasting crash severity levels on the data under test.

A Clustering Regression Approach to Explore the Heterogeneous Effects of Risk Factors Associated with Teen Driver Crash Severity

Teen drivers remain one of the most long-standing traffic safety concerns, as they continue to be overrepresented in the fatal and injury crash statistics. No previous studies have explored significant variations in the effect of risk attributes contributing to teen driver collisions in multiple crash circumstances by degrees of crash severity. Therefore, substantial exploration of teen driver crashes is expected to facilitate the strategic employment of countermeasures effectively. This study aimed to analyze teen driver crashes to investigate the heterogeneous effect of contributing factors on crash severity outcomes. Three years (2017 to 2019) of police-investigated crash information was used for the state of Alabama. This research first applied latent class clustering to minimize the heterogeneity in the extracted dataset by dividing the data into meaningful clusters (subgroups of the whole data). Then, multinomial logit models were constructed to illustrate the significant risk factors influencing the severity outcomes in different crash scenarios. Marginal effects were computed to understand the impacts of variable categories better. The findings suggested that the significance and estimated impact of variables varied within clusters and between crash severity levels in the same cluster. The results of latent class segmented submodels represented real-world crash patterns demonstrating the cumulative effect of variable attributes. Such contextual understandings of underlying risk factors could help to strengthen existing teen driver educational interventions. In addition, the study outcomes could assist practitioners and policy makers in developing safety improvement strategies to reduce the causalities associated with teen driver crashes in distinct circumstances.

Impact of COVID-19 on traffic safety from the “Lockdown” to the “New Normal”: A case study of Utah

During the past several years, the COVID-19 pandemic has had pronounced impacts on traffic safety. Existing studies found that the crash frequency was reduced and the severity level was increased during the earlier “Lockdown” period. However, there is a lack of studies investigating its impacts on traffic safety during the later stage of the pandemic. To bridge such a gap, this study selects Salt Lake County, Utah as the study area and employs statistical methods to investigate whether the impact of COVID-19 on traffic safety differs among different stages. Negative binomial models and binary logit models were utilized to study the effects of the pandemic on the crash frequency and severity respectively while accounting for the exposure, environmental, and human factors. Welch’s t-test and Pairwise t-test are employed to investigate the possible indirect effect of the pandemic by influencing other non-pandemic-related factors in the statistical models. The results show that the crash frequency is significantly less than that of the pre-pandemic during the whole course of the pandemic. However, it significantly increases during the later stage due to the relaxed restrictions. Crash severity levels were increased during the earlier pandemic due to the increased traffic speed, the prevalence of DUI, reduced use of seat belts, and increased presence of commercial vehicles. It reduced to a level comparable to the pre-pandemic later, owing to the reduction of speed and increased seat-belt-used to the pre-pandemic level. As for the incoming “New Normal” stage, stakeholders may need to take actions to deter DUI and reduce commercial-vehicle-related crashes to improve traffic safety.

Elderly Pedestrians and Road Safety: Findings from the Slovenian Accident Database and Measures for Improving Their Safety

The number of elderly people is increasing worldwide, especially in Europe. Such an aging of the population has numerous consequences for society, many of which relate to transportation: older people, aware of their reduced abilities, prefer walking to driving. This leads to an increase in the elderly walking population and thus the need to understand and address the safety issues of these road users. Although these issues are well known, this topic has been little researched so far. The objective of this research is to provide a deeper insight into the safety level of elderly pedestrians by recognizing repetitive patterns leading to accidents involving them, to highlight the magnitude of the problem by analyzing a 10-year pedestrian crash database, to develop a model predicting—on the basis of the recognized patterns—the severity level of collisions involving older pedestrians, and, finally, on the basis of the highlighted factors, to propose some countermeasures to improve their safety. In order to achieve this goal, first, a statistical analysis of the database is performed, considering 13 factors that lead to accidents. Second, Kolmogorov–Smirnov and Anderson–Darling tests are performed to check if the data follow a normal distribution. Finally, an ordinal logistic regression model is proposed to determine the relationship between the crash severity level and the factors characterizing collisions. Thanks to this model, the statistical influencing factors are highlighted. Finally, based on the previous analysis, some technical and educational countermeasures are proposed.

Vehicle crash mitigation strategy in unavoidable collision scenarios: focusing on motion planning by considering a generalized crash severity model

Driving strategy in dynamic environment is crucial to the automated vehicle safety. In extremely emergency scenarios with unavoidable collision (UC), especially those with complex impact patterns, the potential crash risk should be well considered. This paper proposes a crash mitigation (CM) algorithm for UCs, which directly embeds a generalized crash severity index (CSI) model to vehicle-to-vehicle collisions of multiple impact patterns. The idea is that during the short time before a collision, the vehicle will actively adapt its position and poses to minimize the potential crash severity level after the collision. To this end, the generalized CSI model is introduced to estimate the potential crash severity of all sample paths, from which a crash-severity-optimal trajectory is obtained. To improve the inferring time efficiency of the planning module, a neural network is constructed and deployed to approximate the nonlinear severity model. The proposed algorithm is first validated through simulations of UC scenarios, including entry ramp merging, intersection crossing and downhill/uphill crossing. Then for the intersection crossing scenario, the algorithm is deployed to a real car and validated through digital-twin experiments. Results show that by combining the braking and steering interventions for better crash severity reduction, the proposed strategy can achieve better mitigation effects than commonly used collision avoidance (CA) strategies. This reveals that a new mindset of comprehensive safety strategy should not focus only on CA, but also the last resort of CM if collision is unavoidable. Our work may contribute as a promising solution to the safety problem in emergency scenarios.

Modeling the Impact of Driving Styles on Crash Severity Level Using SHRP 2 Naturalistic Driving Data

Previous studies have examined driving styles and how they are associated with crash risks relying on self-report questionnaires to categorize respondents based on pre-defined driving styles. Naturalistic driving studies provide a unique opportunity to examine this relationship differently. The current study aimed to study how driving styles, derived from real-road driving, may relate to crash severity. To study the relationship, this study retrieved safety critical events (SCEs) from the SHRP 2 database and adopted joint modelling of the number of the aggregated crash severity levels (crash vs. non-crash) using the Diagonal Inflated Bivariate Poisson (DIBP) model. Variables examined included driving styles and various driver characteristics. Among driving styles examined, styles of maintenance of lower speeds and more adaptive responses to driving conditions were associated with fewer crashes given an SCE occurred. Longer driving experiences, more miles driven last year, and being female also reduced the number of crashes. Interestingly, older drivers were associated with both an increased number of crashes and increased number of non-crash SCEs. Future work may leverage more variables from the SHRP 2 database and widen the scope to examine different traffic conditions for a more complete picture of driving styles.

Addressing endogeneity in modeling speed enforcement, crash risk and crash severity simultaneously

Speeding is one of the major significant causes of high crash risk and the associated injury severity outcomes. To combat such significant safety concerns, a speed limit enforcement system has been adopted widely around the world. This study aims to present an econometric approach that estimates the casual effect of speed enforcement on safety while addressing the endogeneity issue by employing an instrumental variable approach within a maximum simulated likelihood framework. In our study, safety enforcement is represented as the number of speeding tickets issued from the speed camera systems, while safety profile is presented as two dimensions of interest, including total crash risk and crashes by injury severity levels. The proposed econometric model takes the form of a correlated panel random parameters model with speed enforcement endogeneity. In estimating the joint panel model, speed enforcement and crash severity components are modeled by employing Random Parameters Ordered Logit Fractional Split technique, while ‘ is modeled by employing Random Parameters Negative Binomial regression technique. In the current study context, the ‘operational duration of speed camera’ serves as the instrumental variable for controlling the endogeneity between speed enforcement and safety. Further, the analysis is augmented by a detailed policy scenario analysis. The empirical analysis is demonstrated by employing roadway segment-level crash data and speeding tickets data from Queensland, Australia, for the years 2010 through 2013. From the policy analysis, it is found that a stricter speed enforcement for serious level of speeding offenses is likely to have greater safety benefits in reducing crash severity levels. Moreover, a targeted increase in operation duration along with stricter citations for major speeding is likely to have significant safety gain. The outcome of the study will allow the decision-makers to identify a robust resource allocation and speed camera deployment plan.

Prediction of Crash Severity as a Way of Road Safety Improvement: The Case of Saint Petersburg, Russia

This article investigates factors that explain road crash severity levels in Saint Petersburg, Russia, during the 2015–2021 period. The research takes into account factors such as lighting conditions, weather conditions, infrastructure factors, human factors, accident types, and vehicle category and color to assess their influence on crash severity. The most influential accident type is run-off-road crashes, which are associated with an 11.2% increase in fatal accidents. The biggest reason for the increase in fatal accidents due to road infrastructure conditions is road barrier shortcomings (2.8%). Road infrastructure conditions, such as a lack of road lighting, have a significant effect on fatal outcomes, increasing them by 12.6%, and this is the most influential factor in the analysis. The obtained results may serve as a basis for Saint Petersburg authorities to develop new road safety policies.

Cross-city crash severity analysis with cost-sensitive transfer learning algorithm

Due to the difficulties of data collection in some cities and the relatively small portion of more severe crashes, this paper proposes a cost-sensitive transfer learning framework for more robust crash severity analysis. Specifically, to address the class imbalance issue, cost-sensitive learning method is adopted by assigning unequal cost values to different crash severity levels to obtain unbiased classification results. Moreover, due to the existence of city-irrelevant common crash contributing factors, cross-city crash severity analysis is developed, i.e., transferring common crash severity knowledge from other cities with transfer learning algorithm to assist crash severity modelling in target city that suffers from data scarcity problem. In this study, the crash datasets of Victorian Australia and Seattle are selected as the target and source domains respectively. To address the issue of heterogeneous explanatory features among these two datasets and extract the interpretable common crash severity knowledge, a feature alignment approach is proposed which can represent the cross-city data with unified feature representations on their original explanatory feature spaces. The two proposed models, i.e., cost-sensitive transfer logistic regression (CST-LR) and cost-sensitive transfer support vector machine (CST-SVM), have demonstrated better performance in comparison with twelve commonly used crash severity models, especially when target city crash data is scarce. The most significant crash contributing factors extracted by proposed model also show higher degree of consistency with the true contributing factors obtained from the target dataset, in comparison with the model built without transfer learning. The results could provide policy implications and counter measures for crash severity mitigation.