Analysis of traffic accidents on rural highways using Latent Class Clustering and Bayesian Networks

One of the principal objectives of traffic accident analyses is to identify key factors that affect the severity of an accident. However, with the presence of heterogeneity in the raw data used, the analysis of traffic accidents becomes difficult. In this paper, Latent Class Cluster (LCC) is used as a preliminary tool for segmentation of 3229 accidents on rural highways in Granada (Spain) between 2005 and 2008. Next, Bayesian Networks (BNs) are used to identify the main factors involved in accident severity for both, the entire database (EDB) and the clusters previously obtained by LCC. The results of these cluster-based analyses are compared with the results of a full-data analysis. The results show that the combined use of both techniques is very interesting as it reveals further information that would not have been obtained without prior segmentation of the data. BN inference is used to obtain the variables that best identify accidents with killed or seriously injured. Accident type and sight distance have been identify in all the cases analysed; other variables such as time, occupant involved or age are identified in EDB and only in one cluster; whereas variables vehicles involved, number of injuries, atmospheric factors, pavement markings and pavement width are identified only in one cluster.     

Identifying significant predictors of injury severity in traffic accidents using a series of artificial neural networks

Understanding the circumstances under which drivers and passengers are more likely to be killed or more severely injured in an automobile accident can help improve the overall driving safety situation. Factors that affect the risk of increased injury of occupants in the event of an automotive accident include demographic or behavioral characteristics of the person, environmental factors and roadway conditions at the time of the accident occurrence, technical characteristics of the vehicle itself, among others. This study uses a series of artificial neural networks to model the potentially non-linear relationships between the injury severity levels and crash-related factors. It then conducts sensitivity analysis on the trained neural network models to identify the prioritized importance of crash-related factors as they apply to different injury severity levels. In the process, the problem of five-class prediction is decomposed into a set of binary prediction models (using a nationally representative sample of 30358 police-recorded crash reports) in order to obtain the granularity of information needed to identify the "true" cause and effect relationships between the crash-related factors and different levels of injury severity. The results, mostly validated by the findings of previous studies, provide insight into the changing importance of crash factors with the changing injury severity levels.     

Bayesian networks for maritime traffic accident prevention: Benefits and challenges

Bayesian networks are quantitative modeling tools whose applications to the maritime traffic safety context are becoming more popular. This paper discusses the utilization of Bayesian networks in maritime safety modeling. Based on literature and the author's own experiences, the paper studies what Bayesian networks can offer to maritime accident prevention and safety modeling and discusses a few challenges in their application to this context. It is argued that the capability of representing rather complex, not necessarily causal but uncertain relationships makes Bayesian networks an attractive modeling tool for the maritime safety and accidents. Furthermore, as the maritime accident and safety data is still rather scarce and has some quality problems, the possibility to combine data with expert knowledge and the easy way of updating the model after acquiring more evidence further enhance their feasibility. However, eliciting the probabilities from the maritime experts might be challenging and the model validation can be tricky. It is concluded that with the utilization of several data sources, Bayesian updating, dynamic modeling, and hidden nodes for latent variables, Bayesian networks are rather well-suited tools for the maritime safety management and decision-making.     

Mixed logit model-based driver injury severity investigations in single- and multi-vehicle crashes on rural two-lane highways

Crashes occurring on rural two-lane highways are more likely to result in severe driver incapacitating injuries and fatalities. In this study, mixed logit models are developed to analyze driver injury severities in single-vehicle (SV) and multi-vehicle (MV) crashes on rural two-lane highways in New Mexico from 2010 to 2011. A series of significant contributing factors in terms of driver behavior, weather conditions, environmental characteristics, roadway geometric features and traffic compositions, are identified and their impacts on injury severities are quantified for these two types of crashes, respectively. Elasticity analyses and transferability tests were conducted to better understand the models’ specification and generality. The research findings indicate that there are significant differences in causal attributes determining driver injury severities between SV and MV crashes. For example, more severe driver injuries and fatalities can be observed in MV crashes when motorcycles or trucks are involved. Dark lighting conditions and dusty weather conditions are found to significantly increase MV crash injury severities. However, SV crashes demonstrate different characteristics influencing driver injury severities. For example, the probability of having severe injury outcomes is higher when vans are identified in SV crashes. Drivers’ overtaking actions will significantly increase SV crash injury severities. Although some common attributes, such as alcohol impaired driving, are significant in both SV and MV crash severity models, their effects on different injury outcomes vary substantially. This study provides a better understanding of similarities and differences in significant contributing factors and their impacts on driver injury severities between SV and MV crashes on rural two-lane highways. It is also helpful to develop cost-effective solutions or appropriate injury prevention strategies for rural SV and MV crashes.     

Analysis of driver injury severity in wrong-way driving crashes on controlled-access highways

For more than five decades, wrong-way driving (WWD) has been notorious as a traffic safety issue for controlled-access highways. Numerous studies and efforts have tried to identify factors that contribute to WWD occurrences at these sites in order to delineate between WWD and non-WWD crashes. However, none of the studies investigate the effect of various confounding variables on the injury severity being sustained by the at-fault drivers in a WWD crash. This study tries to fill this gap in the existing literature by considering possible variables and taking into account the ordinal nature of injury severity using three different ordered-response models: ordered logit or proportional odds (PO), generalized ordered logit (GOL), and partial proportional odds (PPO) model. The findings of this study reveal that a set of variables, including driver’s age, condition (i.e., intoxication), seatbelt use, time of day, airbag deployment, type of setting, surface condition, lighting condition, and type of crash, has a significant effect on the severity of a WWD crash. Additionally, a comparison was made between the three proposed methods. The results corroborate that the PPO model outperforms the other two models in terms of modeling injury severity using our database. Based on the findings, several countermeasures at the engineering, education, and enforcement levels are recommended.     

Analysis of traffic injury severity: an application of non-parametric classification tree techniques

Statistical regression models, such as logit or ordered probit/logit models, have been widely employed to analyze injury severity of traffic accidents. However, most regression models have their own model assumptions and pre-defined underlying relationships between dependent and independent variables. If these assumptions are violated, the model could lead to erroneous estimations of injury likelihood. The classification and regression tree (CART), one of the most widely applied data mining techniques, has been commonly employed in business administration, industry, and engineering. CART does not require any pre-defined underlying relationship between target (dependent) variable and predictors (independent variables) and has been shown to be a powerful tool, particularly for dealing with prediction and classification problems. This study uses the 2001 accident data for Taipei, Taiwan. A CART model was developed to establish the relationship between injury severity and driver/vehicle characteristics, highway/environmental variables and accident variables. The results indicate that the most important variable associated with crash severity is the vehicle type. Pedestrians, motorcycle and bicycle riders are identified to have higher risks of being injured than other types of vehicle drivers in traffic accidents.     

Driver injury severity outcome analysis in rural interstate highway crashes: a two-level Bayesian logistic regression interpretation

There is a high potential of severe injury outcomes in traffic crashes on rural interstate highways due to the significant amount of high speed traffic on these corridors. Hierarchical Bayesian models are capable of incorporating between-crash variance and within-crash correlations into traffic crash data analysis and are increasingly utilized in traffic crash severity analysis. This paper applies a hierarchical Bayesian logistic model to examine the significant factors at crash and vehicle/driver levels and their heterogeneous impacts on driver injury severity in rural interstate highway crashes. Analysis results indicate that the majority of the total variance is induced by the between-crash variance, showing the appropriateness of the utilized hierarchical modeling approach. Three crash-level variables and six vehicle/driver-level variables are found significant in predicting driver injury severities: road curve, maximum vehicle damage in a crash, number of vehicles in a crash, wet road surface, vehicle type, driver age, driver gender, driver seatbelt use and driver alcohol or drug involvement. Among these variables, road curve, functional and disabled vehicle damage in crash, single-vehicle crashes, female drivers, senior drivers, motorcycles and driver alcohol or drug involvement tend to increase the odds of drivers being incapably injured or killed in rural interstate crashes, while wet road surface, male drivers and driver seatbelt use are more likely to decrease the probability of severe driver injuries. The developed methodology and estimation results provide insightful understanding of the internal mechanism of rural interstate crashes and beneficial references for developing effective countermeasures for rural interstate crash prevention.     

Underreporting in traffic accident data, bias in parameters and the structure of injury severity models

Injury severities in traffic accidents are usually recorded on ordinal scales, and statistical models have been applied to investigate the effects of driver factors, vehicle characteristics, road geometrics and environmental conditions on injury severity. The unknown parameters in the models are in general estimated assuming random sampling from the population. Traffic accident data however suffer from underreporting effects, especially for lower injury severities. As a result, traffic accident data can be regarded as outcome-based samples with unknown population shares of the injury severities. An outcome-based sample is overrepresented by accidents of higher severities. As a result, outcome-based samples result in biased parameters which skew our inferences on the effect of key safety variables such as safety belt usage. The pseudo-likelihood function for the case with unknown population shares, which is the same as the conditional maximum likelihood for the case with known population shares, is applied in this study to examine the effects of severity underreporting on the parameter estimates. Sequential binary probit models and ordered-response probit models of injury severity are developed and compared in this study. Sequential binary probit models assume that the factors determining the severity change according to the level of the severity itself, while ordered-response probit models assume that the same factors correlate across all levels of severity. Estimation results suggest that the sequential binary probit models outperform the ordered-response probit models, and that the coefficient estimates for lap and shoulder belt use are biased if underreporting is not considered. Mean parameter bias due to underreporting can be significant. The findings show that underreporting on the outcome dimension may induce bias in inferences on a variety of factors. In particular, if underreporting is not accounted for, the marginal impacts of a variety of factors appear to be overestimated. Fixed objects and environmental conditions are overestimated in their impact on injury severity, as is the effect of separate lap and shoulder belt use. Combined lap and shoulder belt usage appears to be unaffected. The parameter bias is most pronounced when underreporting of possible injury accidents in addition to property damage only accidents is taken into account.     

Analysis of injury severity of drivers involved in single- and two-vehicle crashes on highways in Ontario

This study analyzes driver's injury severity in single- and two-vehicle crashes and compares the effects of explanatory variables among various types of crashes. The study identified factors affecting injury severity and their effects on severity levels using 5-year crash records for provincial highways in Ontario, Canada. Considering heteroscedasticity in the effects of explanatory variables on injury severity, the heteroscedastic ordered logit (HOL) models were developed for single- and two-vehicle crashes separately. The results of the models show that there exists heteroscedasticity for young drivers (≤30), safety equipment and ejection in the single-vehicle crash model, and female drivers, safety equipment and head-on collision in the two-vehicle crash models. The results also show that young car drivers have opposite effects between single-car and car–car crashes, and sideswipe crashes have opposite effects between car–car and truck–truck crashes. The study demonstrates that separate HOL models for single-vehicle and different types of two-vehicle crashes can identify differential effects of factors on driver's injury severity.     

Analysis of driver injury severity in rural single-vehicle crashes

Rural roads carry less than fifty percent of the traffic in the United States. However, more than half of the traffic accident fatalities occurred on rural roads. This research focuses on analyzing injury severities involving single-vehicle crashes on rural roads, utilizing a latent class logit (LCL) model. Similar to multinomial logit (MNL) models, the LCL model has the advantage of not restricting the coefficients of each explanatory variable in different severity functions to be the same, making it possible to identify the impacts of the same explanatory variable on different injury outcomes. In addition, its unique model structure allows the LCL model to better address issues pertinent to the independence from irrelevant alternatives (IIA) property. A MNL model is also included as the benchmark simply because of its popularity in injury severity modeling. The model fitting results of the MNL and LCL models are presented and discussed. Key injury severity impact factors are identified for rural single-vehicle crashes. Also, a comparison of the model fitting, analysis marginal effects, and prediction performance of the MNL and LCL models are conducted, suggesting that the LCL model may be another viable modeling alternative for crash-severity analysis.     

Examining driver injury severity outcomes in rural non-interstate roadway crashes using a hierarchical ordered logit model

Rural non-interstate crashes induce a significant amount of severe injuries and fatalities. Examination of such injury patterns and the associated contributing factors is of practical importance. Taking into account the ordinal nature of injury severity levels and the hierarchical feature of crash data, this study employs a hierarchical ordered logit model to examine the significant factors in predicting driver injury severities in rural non-interstate crashes based on two-year New Mexico crash records. Bayesian inference is utilized in model estimation procedure and 95% Bayesian Credible Interval (BCI) is applied to testing variable significance. An ordinary ordered logit model omitting the between-crash variance effect is evaluated as well for model performance comparison. Results indicate that the model employed in this study outperforms ordinary ordered logit model in model fit and parameter estimation. Variables regarding crash features, environment conditions, and driver and vehicle characteristics are found to have significant influence on the predictions of driver injury severities in rural non-interstate crashes. Factors such as road segments far from intersection, wet road surface condition, collision with animals, heavy vehicle drivers, male drivers and driver seatbelt used tend to induce less severe driver injury outcomes than the factors such as multiple-vehicle crashes, severe vehicle damage in a crash, motorcyclists, females, senior drivers, driver with alcohol or drug impairment, and other major collision types. Research limitations regarding crash data and model assumptions are also discussed. Overall, this research provides reasonable results and insight in developing effective road safety measures for crash injury severity reduction and prevention.     

Risk factors associated with traffic violations and accident severity in China

With the recent economic boom in China, vehicle volume and the number of traffic accident fatalities have become the highest in the world. Meanwhile, traffic accidents have become the leading cause of death in China. Systematically analyzing road safety data from different perspectives and applying empirical methods/implementing proper measures to reduce the fatality rate will be an urgent and challenging task for China in the coming years. In this study, we analyze the traffic accident data for the period 2006–2010 in Guangdong Province, China. These data, extracted from the Traffic Management Sector-Specific Incident Case Data Report, are the only officially available and reliable source of traffic accident data (with a sample size >7000 per year). In particular, we focus on two outcome measures: traffic violations and accident severity. Human, vehicle, road and environmental risk factors are considered. First, the results establish the role of traffic violations as one of the major risks threatening road safety. An immediate implication is: if the traffic violation rate could be reduced or controlled successfully, then the rate of serious injuries and fatalities would be reduced accordingly. Second, specific risk factors associated with traffic violations and accident severity are determined. Accordingly, to reduce traffic accident incidence and fatality rates, measures such as traffic regulations and legislation—targeting different vehicle types/driver groups with respect to the various human, vehicle and environment risk factors—are needed. Such measures could include road safety programs for targeted driver groups, focused enforcement of traffic regulations and road/transport facility improvements. Data analysis results arising from this study will shed lights on the development of similar (adjusted) measures to reduce traffic violations and/or accident fatalities and injuries, and to promote road safety in other regions.     

Modelling dependable systems using hybrid Bayesian networks

A hybrid Bayesian network (BN) is one that incorporates both discrete and continuous nodes. In our extensive applications of BNs for system dependability assessment, the models are invariably hybrid and the need for efficient and accurate computation is paramount. We apply a new iterative algorithm that efficiently combines dynamic discretisation with robust propagation algorithms on junction tree structures to perform inference in hybrid BNs. We illustrate its use in the field of dependability with two example of reliability estimation. Firstly we estimate the reliability of a simple single system and next we implement a hierarchical Bayesian model. In the hierarchical model we compute the reliability of two unknown subsystems from data collected on historically similar subsystems and then input the result into a reliability block model to compute system level reliability. We conclude that dynamic discretisation can be used as an alternative to analytical or Monte Carlo methods with high precision and can be applied to a wide range of dependability problems.     

Using hierarchical Bayesian binary probit models to analyze crash injury severity on high speed facilities with real-time traffic data

Severe crashes are causing serious social and economic loss, and because of this, reducing crash injury severity has become one of the key objectives of the high speed facilities’ (freeway and expressway) management. Traditional crash injury severity analysis utilized data mainly from crash reports concerning the crash occurrence information, drivers’ characteristics and roadway geometric related variables. In this study, real-time traffic and weather data were introduced to analyze the crash injury severity. The space mean speeds captured by the Automatic Vehicle Identification (AVI) system on the two roadways were used as explanatory variables in this study; and data from a mountainous freeway (I-70 in Colorado) and an urban expressway (State Road 408 in Orlando) have been used to identify the analysis result's consistence. Binary probit (BP) models were estimated to classify the non-severe (property damage only) crashes and severe (injury and fatality) crashes. Firstly, Bayesian BP models’ results were compared to the results from Maximum Likelihood Estimation BP models and it was concluded that Bayesian inference was superior with more significant variables. Then different levels of hierarchical Bayesian BP models were developed with random effects accounting for the unobserved heterogeneity at segment level and crash individual level, respectively. Modeling results from both studied locations demonstrate that large variations of speed prior to the crash occurrence would increase the likelihood of severe crash occurrence. Moreover, with considering unobserved heterogeneity in the Bayesian BP models, the model goodness-of-fit has improved substantially. Finally, possible future applications of the model results and the hierarchical Bayesian probit models were discussed.     

Bayes classifiers for imbalanced traffic accidents datasets

Traffic accidents data sets are usually imbalanced, where the number of instances classified under the killed or severe injuries class (minority) is much lower than those classified under the slight injuries class (majority). This, however, supposes a challenging problem for classification algorithms and may cause obtaining a model that well cover the slight injuries instances whereas the killed or severe injuries instances are misclassified frequently. Based on traffic accidents data collected on urban and suburban roads in Jordan for three years (2009–2011); three different data balancing techniques were used: under-sampling which removes some instances of the majority class, oversampling which creates new instances of the minority class and a mix technique that combines both. In addition, different Bayes classifiers were compared for the different imbalanced and balanced data sets: Averaged One-Dependence Estimators, Weightily Average One-Dependence Estimators, and Bayesian networks in order to identify factors that affect the severity of an accident. The results indicated that using the balanced data sets, especially those created using oversampling techniques, with Bayesian networks improved classifying a traffic accident according to its severity and reduced the misclassification of killed and severe injuries instances. On the other hand, the following variables were found to contribute to the occurrence of a killed causality or a severe injury in a traffic accident: number of vehicles involved, accident pattern, number of directions, accident type, lighting, surface condition, and speed limit. This work, to the knowledge of the authors, is the first that aims at analyzing historical data records for traffic accidents occurring in Jordan and the first to apply balancing techniques to analyze injury severity of traffic accidents.     

A multinomial logit analysis of risk factors influencing road traffic injury severities in the Erzurum and Kars Provinces of Turkey

A retrospective cross-sectional study is conducted analysing 11,771 traffic accidents reported by the police between January 2008 and December 2013 which are classified into three injury severity categories: fatal, injury, and no injury. Based on this classification, a multinomial logit analysis is performed to determine the risk factors affecting the severity of traffic injuries. The estimation results reveal that the following factors increase the probability of fatal injuries: drivers over the age of 65; primary-educated drivers; single-vehicle accidents; accidents occurring on state routes, highways or provincial roads; and the presence of pedestrian crosswalks. The results also indicate that accidents involving cars or private vehicles or those occurring during the evening peak, under clear weather conditions, on local city streets or in the presence of traffic lights decrease the probability of fatal injuries. This study comprises the most comprehensive database ever created for a Turkish sample. This study is also the first attempt to use an unordered response model to determine risk factors influencing the severity of traffic injuries in Turkey.     

Injury severities of truck drivers in single- and multi-vehicle accidents on rural highways

In adverse driving conditions, such as inclement weather and/or complex terrain, trucks are often involved in single-vehicle (SV) accidents in addition to multi-vehicle (MV) accidents. Ten-year accident data involving trucks on rural highway from the Highway Safety Information System (HSIS) is studied to investigate the difference in driver-injury severity between SV and MV accidents by using mixed logit models. Injury severity from SV and MV accidents involving trucks on rural highways is modeled separately and their respective critical risk factors such as driver, vehicle, temporal, roadway, environmental and accident characteristics are evaluated. It is found that there exists substantial difference between the impacts from a variety of variables on the driver-injury severity in MV and SV accidents. By conducting the injury severity study for MV and SV accidents involving trucks separately, some new or more comprehensive observations, which have not been covered in the existing studies can be made. Estimation findings indicate that the snow road surface and light traffic indicators will be better modeled as random parameters in SV and MV models respectively. As a result, the complex interactions of various variables and the nature of truck-driver injury are able to be disclosed in a better way. Based on the improved understanding on the injury severity of truck drivers from truck-involved accidents, it is expected that more rational and effective injury prevention strategy may be developed for truck drivers under different driving conditions in the future.     

Factors influencing injury severity of motor vehicle-crossing pedestrian crashes in rural Connecticut

The ordered probit model was used to evaluate the effect of roadway and area type features on injury severity of pedestrian crashes in rural Connecticut. Injury severity was coded on the KABCO scale and crashes were limited to those in which the pedestrians were attempting to cross two-lane highways that were controlled by neither stop signs nor traffic signals. Variables that significantly influenced pedestrian injury severity were clear roadway width (the distance across the road including lane widths and shoulders, but excluding the area occupied by on-street parking), vehicle type, driver alcohol involvement, pedestrian age 65 years or older, and pedestrian alcohol involvement. Seven area types were identified: downtown, compact residential, village, downtown fringe, medium-density commercial, low-density commercial, and low-density residential. Two groups of these area types were found to experience significantly different injury severities. Downtown, compact residential, and medium- and low-density commercial areas generally experienced lower pedestrian injury severity than village, downtown fringe, and low-density residential areas.     

Traffic accident segmentation by means of latent class clustering

Traffic accident data are often heterogeneous, which can cause certain relationships to remain hidden. Therefore, traffic accident analysis is often performed on a small subset of traffic accidents or several models are built for various traffic accident types. In this paper, we examine the effectiveness of a clustering technique, i.e. latent class clustering, for identifying homogenous traffic accident types. Firstly, a heterogeneous traffic accident data set is segmented into seven clusters, which are translated into seven traffic accident types. Secondly, injury analysis is performed for each cluster. The results of these cluster-based analyses are compared with the results of a full-data analysis. This shows that applying latent class clustering as a preliminary analysis can reveal hidden relationships and can help the domain expert or traffic safety researcher to segment traffic accidents.     

Drug use and the severity of a traffic accident

Several studies have showed that driving under the influence of alcohol and/or certain illicit or medicinal drugs increases the risk of a (severe) crash. Data with respect to the question whether this also leads to a more severe accident are sparse. This study examines the relationship between the use of alcohol, illicit drugs and/or medicinal drugs and the severity of an accident within a group of drivers that were involved in a crash in The Netherlands. Blood samples of 993 drivers, collected in the period from October 1998 through September 1999, were linked to accident characteristics as available from the National Transport Research Centre. The outcome measure was the severity of the accident. An accident was considered severe when the accident had resulted in hospital admission or death. All the blood samples obtained after the accident were screened for the presence of alcohol, illicit drugs (opiates, amphetamines and amphetamine-like substances, cocaine and metabolites, methadone, cannabinoids) and medicinal drugs (benzodiazepines, barbiturates and tricyclic antidepressants). The strength of the associations between exposure to the different classes of alcohol/drugs/medicines and the severity of the accident was evaluated using logistic regression analysis and were expressed as odds ratios (OR), adjusted for age, gender, time of the day, day of the week and urban area. The most frequently detected drugs were cannabinoids, benzodiazepines and cocaine. Our results showed no clear association between the use of alcohol, illicit drug and/or medicinal drug use and the severity of the accident. Given the process of obtaining blood samples from drivers involved in accidents and the retrospective nature of the study, we cannot rule out the occurrence of selection bias. Therefore, our findings need further confirmation.