Traffic Crash Severity Research Articles

Traffic Crash Severity Prediction with Convolutional Neural Network and Long Short-Term Memory: An Automatic Feature Selection Approach

Traffic Crash Severity Prediction with Convolutional Neural Network and Long Short-Term Memory: An Automatic Feature Selection Approach

Urban Road Safety and Crash Severity during the 2020 COVID-19 Pandemic: The Case of Seattle, WA

Teleworking and lockdown measures implemented during the COVID-19 pandemic have resulted in reduced traffic flows and fewer cars at risk of collision on roads. The present study presents an analysis of the pandemic’s impact on traffic safety and crash severity in the city of Seattle, WA. We found increased shares of speeding-related, angle-, and head-on collisions and relatively more collisions occurring during off-peak hours. Results of quasi-difference-in-differences modeling further suggest that collisions occurring during the post-lockdown period have had a higher probability of being fatal or resulting in severe injuries, compared to what would be expected in the absence of the pandemic.

Trivariate Copula for Modeling Barriers Crash Severity, Accounting for Policy Endogeneity

Studies have been implemented in the literature to enhance the safety of traffic barriers by identification of the contributory factors to those crashes. However, almost all those past studies are subject to potential pitfalls of just answering the cause-and-effect question by traditional statistical methods, which fail to account for possible endogeneity. Modeling traffic barrier crash severity with traditional statistical methods might be biased as many parameters such as barrier’s types is endogenous to unseen factors including policy makers’ decisions in allocating those barriers. Not accounting for the possible endogeneity in the dataset and ignoring correlation between regressors’ error terms might result in biased or erroneous coefficients’ estimates. That is especially true in the presence of strong correlations across models’ error terms. Thus, this study was conducted to model barriers’ crash severity by taking into consideration the endogeneity and correlations across the models’ error terms. Here, the trivariate copula-based method was implemented to simultaneously model traffic barrier crash severity, shoulder width and barrier’s types, while accounting for interrelationships across the models’ error terms. The results provide strong evidence of correlations between the unseen factors to the selections of barrier’s types, shoulder width installation, and crash injury levels. For instance, we found in the presence of accounting for endogeneity and correlation between unseen factors, concrete traffic barrier type and higher shoulder width installation are negatively correlated with unseen factors contributing to severe barriers’ crashes That is despite the fact that the observed factors of those predictors were found to have a reverse impact on the severity of barriers’ crashes.

Investigation of Contributing Factors to Traffic Crash Severity in Southeast Texas Using Multiple Correspondence Analysis

Driving is the essential means of travel in Southeast Texas, a highly urbanized and populous area that serves as an economic powerhouse of the whole state. However, driving in Southeast Texas is subject to many risks as this region features a typical humid subtropical climate with long hot summers and short mild winters. Local drivers would encounter intense precipitation, heavy fog, strong sunlight, standing water, slick road surface, and even frequent extreme weather such as tropical storms, hurricanes and flood during their year-around travels. Meanwhile, research has revealed that the fatality rate per 100 million vehicle miles driven in urban Texas became considerably higher than national average since 2010, and no conclusive study has elucidated the association between Southeast Texas crash severity and potential contributing factors. This study used multiple correspondence analysis (MCA) to examine a group of contributing factors on how their combinatorial influences determine crash severity by creating combination clouds on a factor map. Results revealed numerous significant combinatorial effects. For example, driving in rain and extreme weather on a wet road surface has a higher chance in causing crashes that incur severe or deadly injuries. Besides, other contributing factors involving risky behavioral factors, road designs, and vehicle factors were well discussed. The research outcomes could inspire local traffic administration to take more effective countermeasures to systematically mitigate road crash severity.

Diagnostic analysis of environmental factors affecting the severity of traffic crashes: From the perspective of pedestrian–vehicle and vehicle–vehicle collisions

Objectives Traffic crashes under low-visibility conditions are frequent and serious. The aim of this study was to investigate how the road environment affects the severity of pedestrian–vehicle and vehicle–vehicle collisions under low-visibility conditions. Methods The injury severity of pedestrian–vehicle collisions and vehicle–vehicle collisions under low-visibility conditions was set as the dependent variable and divided into 2 categories: “killed or severe injury collision” and “slight injury collision.” Ten variables, including environment conditions, road traffic facility status, collision characteristics, and road attributes, were selected as independent factors according to the existing research and the traffic collision data set. Based on 656 valid pedestrian–vehicle collisions and 1,430 valid vehicle–vehicle collisions under low-visibility conditions, 2 random parameter logit models were established to evaluate the impacts of influencing factors on the severity of pedestrian–vehicle collisions and vehicle–vehicle collisions, in which the effect of unobserved heterogeneity was accounted for. Results The results show that visibility, presence of a roadside protection, road type, road pavement condition, and road alignment were significant factors affecting the severity of pedestrian–vehicle collisions. In addition, the presence of a median divider, location of the collision, road type, road surface condition, road pavement condition, and road alignment were significant factors affecting the severity of vehicle–vehicle collisions. Furthermore, the injury severity of both pedestrian–vehicle collisions and vehicle–vehicle collisions under low-visibility conditions on highways, poor road pavement, and non-straight-line sections was more likely to be fatal or serious. Conclusion These results have implications for the design of more effective strategies to reduce casualties from traffic crashes under low-visibility conditions.

Analysis of Traffic Accident Features and Crash Severity Prediction

Vehicle crashes occur because of numerous factors. It leads to loss of lives and permanent incapacity. The budgetary expenses of both individuals as well as for the nation are influenced by vehicle crashes. According to Road accident statistics, a total of 464910 road accidents were reported in India, claiming 1,47,913 lives and causing injuries to 4,70,975 persons every year. In this work, the UK data set sourced from Kaggle is used. For the study, 17 attributes and 35k records of the year 2015 are considered. The data set is imbalanced, so to balance out the data, the over-sampling technique is used. Random Forest, Decision tree, Logistic Regression, and Gradient Naïve Bayes algorithms are used to predict the severity of Accidents. To evaluate the model, performance measures like Accuracy, Precision, Recall, F1-Score are used. When Accuracy, Precision, F1-Score performance measure is considered Random Forest yielded the best result. When Recall performance measure is used, Random forest for Fatal, Decision Trees for Serious, Logistic regression for Slight yielded the best result.

Organizational and personal factors in occupational traffic injuries at work in Spain

Objective Traffic crashes are one of the major causes of deaths at work. However, most of the literature on traffic crashes did not study the influence of worker and organization conditions. The current paper analyses the influence of the personal and organizational factors associated with the severity of occupational traffic crashes. Methods A total of 724,596 occupational traffic crashes that occurred in Spain from 2009 to 2019 were analyzed. Men suffered 59% of crashes registered, while workers younger than 30 represented 31.3% of crashes studied. Variables included in the Official Occupational Accident Report were analyzed using logistic regression calculating odds ratios. Results Results showed that a foreign, professional male driver from a small company who is older than 55 years was more likely to suffer a fatal crash. In contrast, a Spanish, nonprofessional female driver from a medium or big company who is younger than 30 years is more likely to suffer light consequences. Conclusions Findings from the current research could help to improve the safety training programmes in companies considering their personal and business variables such as age, gender, or company staff, especially in small companies. Future research should collect and analyze additional variables combining different data sources.

Applications of machine learning methods in traffic crash severity modelling: current status and future directions

ABSTRACT As a key area of traffic safety research, crash severity modelling has attracted tremendous attention. Recently, there has been growing interest in applying machine learning (ML) methods in this area. However, the lessons and experience learned so far have not been systematically documented and summarised. This is the first article that surveys studies on ML applications in crash severity modelling and has the following major contributions: (1) it provides a comprehensive and critical review of current research efforts; (2) it summarises the successful experience and main challenges (e.g. data and methodology); and (3) it identifies promising research opportunities towards accurate and reliable crash severity modelling and results interpretation. The review results suggest that imbalanced data remains a major issue. Under- and over-samplings are often used to balance crash severity data despite their limitations. Some studies use local sensitivity analysis (LSA) to interpret ML modelling results but ignore the strict assumptions of LSA and omit the joint effects of risk factors. Moreover, very few studies consider the accuracy and reliability of ML model evaluation metrics. Other issues include spatiotemporal correlations, causality, model transferability and heterogeneity. This paper concludes by providing suggestions on model selection and modification to address the identified issues and recommendations for future research. For example, employing advanced ML methods such as graph convolutional networks (GCN) to model spatiotemporal correlations; exploring innovative ways of applying ML methods; and leveraging new developments in ML (e.g. interpretable ML) to derive causal relationships and interpret modelling results.

Evaluation of fog warning system on driving under heavy fog condition based on driving simulator

Purpose Heavy fog results in low visibility, which increases the probability and severity of traffic crashes, and fog warning system is conducive to the reduction of crashes by conveying warning messages to drivers. This paper aims at exploring the effects of dynamic message sign (DMS) of fog warning system on driver performance. Design/methodology/approach First, a testing platform was established based on driving simulator and driver performance data under DMS were collected. The experiment route was consisted of three different zones (i.e. warning zone, transition zone and heavy fog zone), and mean speed, mean acceleration, mean jerk in the whole zone, ending speed in the warning zone and transition zone, maximum deceleration rate and mean speed reduction proportion in the transition zone and heavy fog zone were selected. Next, the one-way analysis of variance was applied to test the significant difference between the metrics. Besides, drivers’ subjective perception was also considered. Findings The results indicated that DMS is beneficial to reduce speed before drivers enter the heavy fog zone. Besides, when drivers enter a heavy fog zone, DMS can reduce the tension of drivers and make drivers operate more smoothly. Originality/value This paper provides a comprehensive approach for evaluating the effectiveness of the warning system in adverse conditions based on the driving simulation test platform. The method can be extended to the evaluation of vehicle-to-infrastructure technology in other special scenarios.

Analyzing Road Traffic Crash Severity In Uttarakhand Region

Analyzing Road Traffic Crash Severity In Uttarakhand Region

Crash Severity Analysis of Highways Based on Multinomial Logistic Regression Model, Decision Tree Techniques, and Artificial Neural Network: A Modeling Comparison

The classification of vehicular crashes based on their severity is crucial since not all of them have the same financial and injury values. In addition, avoiding crashes by identifying their influential factors is possible via accurate prediction modeling. In crash severity analysis, accurate and time-saving prediction models are necessary for classifying crashes based on their severity. Moreover, statistical models are incapable of identifying the potential severity of crashes regarding influencing factors incorporated in models. Unlike previous research efforts, which focused on the limited class of crash severity, including property damage only (PDO), fatality, and injury by applying data mining models, the present study sought to predict crash frequency according to five severity levels of PDO, fatality, severe injury, other visible injuries, and complaint of pain. The multinomial logistic regression (MLR) model and data mining approaches, including artificial neural network-multilayer perceptron (ANN-MLP) and two decision tree techniques, (i.e., Chi-square automatic interaction detector (CHAID) and C5.0) are utilized based on traffic crash records for State Highways in California, USA. The comparison of the findings of the relative importance of ten qualitative and ten quantitative independent variables incorporated in CHAID and C5.0 indicated that the cause of the crash (X1) and the number of vehicles (X5) were known as the most influential variables involved in the crash. However, the cause of the crash (X1) and weather (X2) were identified as the most contributing variables by the ANN-MLP model. In addition, the MLR model showed that the driver’s age (X11) accounts for a larger proportion of traffic crash severity. Therefore, the sensitivity analysis demonstrated that C5.0 had the best performance for predicting road crash severity. Not only did C5.0 take a shorter time (0.05 s) compared to CHAID, MLP, and MLR, it also represented the highest accuracy rate for the training set. The overall prediction accuracy based on the training data was approximately 88.09% compared to 77.21% and 70.21% for CHAID and MLP models. In general, the findings of this study revealed that C5.0 can be a promising tool for predicting road crash severity.

Spatial analysis and evaluation of road traffic safety performance indexes across the provinces of Turkey from 2015 to 2019

Limited road safety based spatial analysis studies have been conducted in developing countries. Also, little is known about the relationships between province-level road safety performance indexes (RSPIs). Hence, spatial, regression and correlation analysis were used to identify road safety-deficient provinces and determine the relationship between RSPIs. The gathered data comprise 14 RSPIs and nine socioeconomic indicators. Moran’s I and Local Moran indexes were used for conducting the spatial analysis. The natural breaks method was used to cluster similar provinces according to RSPIs. Regarding studied RSPIs, huge local clusters of provinces detected. Eastern provinces had higher road traffic crash (RTC) severity indexes. RTCs were more severe in regions with lower income level. Regions with higher socioeconomic indexes such as population had higher RTC rates. Using RSPIs calculated with distinct exposure measures creates completely different local cluster maps. significant relationships between studied RSPIs were detected. A standard system is needed to organize and categorize the RSPIs. Road safety policies should be region-specific to reduce RSPIs efficiently. Regarding the observed various locations of hot spots in terms of studied RSPIs, further consideration should be given in the process of selecting an RSPI for comparing administrative divisions of a country.

Investigating the Relationship between Traffic Violations and Crashes at Signalized Intersections: An Empirical Study in China

About 90% of traffic crashes are caused by human factors, within which traffic violations are one of the most typical and common causes. In order to investigate the relationship between traffic violations and traffic crashes, this research targets signalized intersections in two Chinese cities: Yinchuan and Suqian. Thirty-one intersections are selected as the research sites, and additionally, the traffic volume, traffic violation, and traffic crash data of each intersection are collected for one year. A White’s test is conducted to test the homoscedasticity of the data and a multiple linear regression model is employed to investigate the relationship between traffic crashes and violations. The results show the following: (1) although the research sites are located in two different cities, the data is homoscedastic, which suggests that the above result may be statistically stable between different cities. (2) There is a significant multiple linear regression relationship (R2 = 0.782, adjusted R2 = 0.716) between the total number of traffic crashes and traffic violations. Among the chosen 7 independent variables, four are significantly related to the dependent variable, namely, driving commercial vehicle during internship, wrong-way entry, speeding, and traffic-light violation. (3) With the increase of annual average daily traffic (AADT), the number of total crashes goes up; however, the injury-or-fatality rate decreases, which means that intersections with smaller traffic volumes tend to have higher traffic crash severity. Based on the above conclusions, it is possible to conduct more targeted enforcement to improve the safety of intersections.

Traffic Crash Severity Prediction with Deep Learning

This paper is based on deep learning to explore the accuracy forecasting model of traffic crash severity. A prediction model in terms of convolutional neural networks is proposed in the paper. The first is to collect the basic information of all traffic crash data of the entire urban road network in Chicago from 2016 to 2020. Then using the PivotTable to observe the distribution of each variable, and filtering them. The required prediction model is based on Python and it mainly includes three parts, the input layer, the hidden layer, and the output layer. Besides, using SGD optimizer to accelerate the training of the network model is a need. Aiming to avoid problems such as over-fitting, the dropout technique is adopted, and its parameter value taking 0.1 is better. Then comparing and analyzing with three traditional machine learning models, the decision tree classification prediction model, the logistic regression prediction model and the support vector machine prediction model, it is obvious to show the better performance of prediction model based on the convolutional neural network method. After optimization, its highest accuracy is 89.2%.

A deep learning based traffic crash severity prediction framework

Highway work zones are most vulnerable roadway segments for congestion and traffic collisions. Hence, providing accurate and timely prediction of the severity of traffic collisions at work zones is vital to reduce the response time for emergency units (e.g., medical aid), accordingly improve traffic safety and reduce congestion. In predicting the severity of traffic collisions, previous studies used different statistical and machine learning models with accuracy as the main evaluating factor. However, the performance of these models was generally not good, especially on fatal and injury crashes. Also, looking into the prediction accuracy only is misleading.This paper aims to propose a novel deep learning-based approach with a customized f1-loss function to predict the severity of traffic crashes. Underlying this objective is to compare the results of deep learning models with machine learning model considering two performance indicators, namely precision, and recall. The data used in the analysis include a sample of traffic crashes that occurred at work zones in Louisiana from 2014 to 2018. This dataset includes valuable information (features) related to road, vehicle, and human factors affecting the occurrence and severity of those crashes.The proposed methodology is based on transforming these features/variables into images. Image transformation is conducted using a nonlinear dimensionality reduction technique t-SNE and convex hull algorithm. A CNN based deep learning algorithm with a customized loss function was used to directly optimize the model for precision and recall. The results showed improved performance in predicting the crash severity of fatal and injury crashes using the deep learning approach, which can help to improve traffic safety as well as traffic congestion at work zones and possibly other roadways segments.

Macro-level hazardous material transportation safety analysis in China using a Bayesian negative binomial model combined with conditional autoregression prior

Traffic safety for hazardous material (hazmat) transportation has not been studied well at a macro level in recent years. A Bayesian negative binomial conditional autoregressive safety model was used within Chinese provinces and cities. A total of 1,229 hazmat transportation crashes in China were collected from the years 2015 to 2017. The frequency of hazmat transportation crashes and the frequency of severe crashes including fatalities and serious injuries were studied in relation to socioeconomic factors, road classification, and the scale of hazmat transportation. The results show that higher crash frequencies are associated with a greater gross domestic product index, increasing road densities, and number of hazmat transportation vehicles and hazmat drivers per vehicle. The frequency of severe crashes tends to be higher in provinces with greater populations, increasing road densities, mileage of low-grade roads, and number of companies. The urban road mileage and number of hazmat loaders are negatively associated with the total number of hazmat crashes and severe crashes. Additionally, the hospital density also has a negative correlation with the frequency of severe traffic crashes. These results could help hazmat transportation managers and planners determine the risk factors of hazmat crashes on a macro level and develop appropriate measures for improving hazmat transportation safety.

Bicycle Rider Behavior and Crash Involvement in Australia.

This research investigated how behaviours and attitudes of bicycle riders influence crash frequency and severity. The study recruited 1102 Australian bicycle riders for an online survey. The survey comprised questions on demographics, frequency of riding and the number and severity of traffic crashes during the last five years. The survey included the Cycling Behaviour Questionnaire and the Cyclist Risk Perception and Regulation Scale. Overall, there were low levels of errors and violations reported by participants indicating that these behaviours were on average never or rarely exhibited while riding a bicycle. Conversely, participants reported high levels of engagement in positive behaviours and reported high levels of traffic rule knowledge and risk perception. Higher rates of violations and errors were associated with increased crash likelihood, while higher rates of positive behaviours were associated with reduced rates of crash involvement in a period of 5 years. The findings highlight the relationship between errors, total crashes and crash severity Further promotion of positive behaviours amongst riders may also help to reduce the risk of crashes.

Older Pedestrian Traffic Crashes Severity Analysis Based on an Emerging Machine Learning XGBoost

Older pedestrians are vulnerable on the streets and at significant risk of injury or death when involved in crashes. Pedestrians’ safety is critical for roadway agencies to consider and improve, especially older pedestrians aged greater than 65 years old. To better protect the older pedestrian group, the factors that contribute to the older crashes need to be analyzed deeply. Traditional modeling approaches such as Logistic models for data analysis may lead to modeling distortions due to the independence assumptions. In this study, Extreme Gradient Boosting (XGBoost), is used to model the classification problem of three different levels of severity of older pedestrian traffic crashes from crash data in Colorado, US. Further, Shapley Additive explanations (SHAP) are implemented to interpret the XGBoost model result and analyze each feature’s importance related to the levels of older pedestrian crashes. The interpretation results show that the driver characteristic, older pedestrian characteristics, and vehicle movement are the most important factors influencing the probability of the three different severity levels. Those results investigate each severity level’s correlation factors, which can inform the department of traffic management and the department of road infrastructure to protect older pedestrians by controlling or managing some of those significant features.

Bayesian hierarchical modelling of traffic barrier crash severity

Severe vehicle crashes have resulted in a large-scale social and economic loss. As a result, the reduction of those crashes has become one of the key objectives of policy makers. Although traffic barriers have been utilized to reduce the run-of-road crash severity, still those crashes account for a high number of severe crashes. Previous studies of traffic barrier crashes often either ignore the heterogeneity across different traffic barrier types, or just focus on specific types of traffic barriers. Thus, this study developed a Bayesian Hierarchical model (BHM) to identify the contributory factors impacting the severity of traffic barrier crashes while accounting for that heterogeneity. The assessment of model fit, inter-class correlation (ICC) coefficient, and deviance information criterion (DIC) all favoured the use of the BHM. Besides accounting for the heterogeneity between barrier types, the interaction across variables shoulder width and traffic barrier height were incorporated into the analysis. Due to the differences in traffic barrier design and vehicle performance across different roadway classifications, only Wyoming interstate traffic barrier were considered. Results indicated that there is an important interaction term between traffic barrier height and shoulder width so that the impact of these two predictors should not be separated. In addition, having a citation record, negotiating a curve, being a female driver, non-speed compliance, alcohol involvement, and showing emotional signs at the time of crash were factors increasing the severity of traffic barrier crashes. On the other hand, having a turn before hitting a traffic barrier, being a younger driver, and driving in adverse weather conditions were factors that significantly decrease the severity of traffic barrier crashes

Collective Risk Ranking of Highway Segments on the Basis of Severity-Weighted Crash Rates

This study is intended to focus on the major factors affecting traffic crash rates and severity levels, in addition to identifying crash-prone locations (i.e., black spots) based on the two indicators. The available crash data for different road segments used for the analysis were obtained from the Washington state database provided by the Highway Safety Information System (HSIS) for the years 2006 to 2011. A Random Forest (RF) classifier was used to predict the outcome level of crash severity, while crash rates were predicted by applying RF regressor. Certain features were selected for each model besides the abstraction of new features to check if there are unobserved correlations affecting the independent variables, such as accounting for the number and weight of crashes within 1 km2 area by implementing the Getis-Ord Gi∗ index. Moreover, to calculate the collective risk (CR) score, crash rates were adjusted to incorporate crash severity weights (cost per severity type) and regression-to-the-mean (RTM) bias via Empirical Bayes (EB) method. Finally, segments were ranked according to their CR score.