Development of crash prediction models with individual vehicular data

Typical engineering research on traffic safety focuses on identifying either dangerous locations or contributing factors through a post-crash analysis using aggregated traffic flow data and crash records. A recent development of transportation engineering technologies provides ample opportunities to enhance freeway traffic safety using individual vehicular information. However, little research has been conducted regarding methodologies to utilize and link such technologies to traffic safety analysis. Moreover, traffic safety research has not benefited from the use of hurdle-type models that might treat excessive zeros more properly than zero-inflated models.This study developed a new surrogate measure, unsafe following condition (UFC), to estimate traffic crash likelihood by using individual vehicular information and applied it to basic sections of interstate highways in Virginia. Individual vehicular data and crash data were used in the development of statistical crash prediction models including hurdle models. The results showed that an aggregated UFC measure was effective in predicting traffic crash occurrence, and the hurdle Poisson model outperformed other count data models in a certain case.     

Geographical unit based analysis in the context of transportation safety planning

A wide array of spatial units has been explored in macro-level modeling. With the advancement of Geographic Information System (GIS) analysts are able to analyze crashes for various geographical units. However, a clear guideline on which geographic entity should be chosen is not present. Macro level safety analysis is at the core of transportation safety planning (TSP) which in turn is a key in many aspects of policy and decision making of safety investments. The preference of spatial unit can vary with the dependent variable of the model. Or, for a specific dependent variable, models may be invariant to multiple spatial units by producing a similar goodness-of-fits. In this study three different crash models were investigated for traffic analysis zones (TAZs), block groups (BGs) and census tracts (CTs) of two counties in Florida. The models were developed for the total crashes, severe crashes and pedestrian crashes in this region. The primary objective of the study was to explore and investigate the effect of zonal variation (scale and zoning) on these specific types of crash models. These models were developed based on various roadway characteristics and census variables (e.g., land use, socio-economic, etc.).It was found that the significance of explanatory variables is not consistent among models based on different zoning systems. Although the difference in variable significance across geographic units was found, the results also show that the sign of the coefficients are reasonable and explainable in all models.Key findings of this study are, first, signs of coefficients are consistent if these variables are significant in models with same response variables, even if geographic units are different. Second, the number of significant variables is affected by response variables and also geographic units.Admittedly, TAZs are now the only traffic related zone system, thus TAZs are being widely used by transportation planners and frequently utilized in research related to macroscopic crash analysis. Nevertheless, considering that TAZs are not delineated for traffic crash analysis but they were designed for the long range transportation plans, TAZs might not be the optimal zone system for traffic crash modeling at the macroscopic level. Therefore, it recommended that other zone systems be explored for crash analysis as well.     

A latent variable representation of count data models to accommodate spatial and temporal dependence: Application to predicting crash frequency at intersections

This paper proposes a reformulation of count models as a special case of generalized ordered-response models in which a single latent continuous variable is partitioned into mutually exclusive intervals. Using this equivalent latent variable-based generalized ordered response framework for count data models, we are then able to gainfully and efficiently introduce temporal and spatial dependencies through the latent continuous variables. Our formulation also allows handling excess zeros in correlated count data, a phenomenon that is commonly found in practice. A composite marginal likelihood inference approach is used to estimate model parameters. The modeling framework is applied to predict crash frequency at urban intersections in Arlington, Texas. The sample is drawn from the Texas Department of Transportation (TxDOT) crash incident files between 2003 and 2009, resulting in 1190 intersection-year observations. The results reveal the presence of intersection-specific time-invariant unobserved components influencing crash propensity and a spatial lag structure to characterize spatial dependence. Roadway configuration, approach roadway functional types, traffic control type, total daily entering traffic volumes and the split of volumes between approaches are all important variables in determining crash frequency at intersections.     

Motorway crash duration and its determinants: do durations vary across motorways?

Traffic congestion has been a growing issue in many metropolitan areas during recent years, which necessitates the identification of its key contributors and development of sustainable strategies to help decrease its adverse impacts on traffic networks. Road incidents generally and crashes specifically have been acknowledged as the cause of a large proportion of travel delays in urban areas and account for 25% to 60% of traffic congestion on motorways. Identifying the critical determinants of travel delays has been of significant importance to the incident management systems, which constantly collect and store the incident duration data. This study investigates the individual and simultaneous differential effects of the relevant determinants on motorway crash duration probabilities. In particular, it applies parametric Accelerated Failure Time (AFT) hazard‐based models to develop in‐depth insights into how the crash‐specific characteristic and the associated temporal and infrastructural determinants impact the duration. AFT models with both fixed and random parameters have been calibrated on one year of traffic crash records from two major Australian motorways in South East Queensland, and the differential effects of determinants on crash survival functions have been studied on these two motorways individually. A comprehensive spectrum of commonly used parametric fixed parameter AFT models, including generalized gamma and generalized F families, has been compared with random parameter AFT structures in terms of goodness of fit to the duration data, and as a result, the random parameter Weibull AFT model has been selected as the most appropriate model. Significant determinants of motorway crash duration included traffic diversion requirement, crash injury type, number and type of vehicles involved in a crash, day of week and time of day, towing support requirement and damage to the infrastructure. A major finding of this research is that the motorways under study are significantly different in terms of crash durations; such that motorway 1 exhibits durations that are on average 19% shorter compared with the durations on motorway 2. The differential effects of explanatory variables on crash durations are also different on the two motorways. The detailed presented analysis confirms that looking at the motorway network as a whole, neglecting the individual differences between roads, can lead to erroneous interpretations of duration and inefficient strategies for mitigating travel delays along a particular motorway.     

Modeling unobserved heterogeneity using finite mixture random parameters for spatially correlated discrete count data

Road segments with identical site-specific attributes often exhibit significantly different crash counts due to unobserved reasons. The extent of unobserved heterogeneity associated with a road feature is to be estimated prior to selecting the relevant safety treatment. Moreover, crash count data is often over-dispersed and spatially correlated. This paper proposes a spatial negative binomial specification with random parameters for modeling crash counts of contiguous road segments. The unobserved heterogeneity is incorporated using a finite multi-variate normal mixture prior on the random parameters; this allows for non-normality, skewness in the distribution of the random parameters, facilitates correlation across the random parameters, and relaxes any distributional assumptions. The model extracts the inherent groups of road segments with crash counts that are equally sensitive to the road attributes on an average; the heterogeneity within these groups is also allowed in the proposed framework. The specification simultaneously accounts for potential spatial correlation of the crash counts from neighboring road segments. A Gibbs sampling framework is proposed that leverages recent theoretical developments on data-augmentation algorithms, and elegantly sidesteps many of the computational difficulties usually associated with Bayesian inference of count models. Empirical results suggests the presence of two latent groups and spatial correlation within the study road network. Road features with significantly different effect on crash counts across two latent groups of road segments were identified.     

Real‐time identification of traffic conditions prone to injury and non‐injury crashes on freeways using genetic programming

This study applied the genetic programming (GP) model to identify traffic conditions prone to injury and property‐damage‐only (PDO) crashes in different traffic states on freeways. It was found that the traffic conditions prone to injury and PDO crashes can be classified into a high‐speed and a low‐speed traffic state. The random forest (RF) analyses were conducted to identify the contributing factors to injury and PDO crashes in these two traffic states. Four separate GP models were then developed to link the risks of injury and PDO crashes in two traffic states to the variables selected by the RF. An overall GP model was also developed for the combined dataset. It was found that the separate GP models that considered different traffic states and crash severity provided better predictive performance than the overall model, and the traffic flow variables that affected injury and PDO crashes were quite different across different traffic states. The proposed GP models were also compared with the traditional logistic regression models. The results suggested that the GP models outperformed the logistic regression models in terms of the prediction accuracy. More specifically, the GP models increased the prediction accuracy of injury crashes by 10.7% and 8.0% in the low‐speed and high‐speed traffic states. For PDO crashes, the GP models increased the prediction accuracy by 7.4% and 6.0% in the low‐speed and high‐speed traffic states. Copyright © 2016 John Wiley & Sons, Ltd.     

A mixed Bayesian network for two-dimensional decision modeling of departure time and mode choice

The modeling of travel decision making has been a popular topic in transportation planning. Previous studies focused on random-utility discrete choice models and machine learning methods. This paper proposes a new modeling approach that utilizes a mixed Bayesian network (BN) for travel decision inference. The authors use a predetermined BN structure and calculate priori and posterior probability distributions of the decision alternatives based on the observed explanatory variables. As a “utility-free” decision inference method, the BN model releases the linear structure in the utility function but assumes the traffic level of service variables follow multivariate Gaussian distribution conditional on the choice variable. A real-world case study is conducted by using the regional travel survey data for a two-dimensional decision modeling of both departure time choice and travel mode choice. The results indicate that a two-dimensional mixed BN provides better accuracy than decision tree models and nested logit models. In addition, one can derive continuous elasticity with respect to each continuous explanatory variable for sensitivity analysis. This new approach addresses a research gap in probabilistic travel decision making modeling as well as two-dimensional travel decision modeling.     

Safety modeling for two‐way left‐turn lanes and a procedure to identify two‐way left‐turn lane segments with safety concerns

In the past, two‐way left‐turn lane (TWLTL) median treatments have been frequently used in Florida to inexpensively improve traffic and safety performances. In order to identify factors that may have significant impacts on safety operations in TWLTL sections and to identify TWLTL locations that present existing and future safety concerns, a research project was carried out and results are summarized in the paper. In the research, a three‐year crash history database with crashes and section characteristics from a total of 1688 TWLTL sections all over Florida was developed and used. A negative binomial regression model was developed to determine the statistical relationship between the number of crashes per mile per year and several variables such as traffic volume, access density, posted speed, and number of lanes. In regard to the methodology, in order to identify locations with safety concerns, several steps are needed: development of real crash data distribution, determination of statistical distribution models that better represent the actual crash data, determination of percentile values for the average number of crashes, estimation of crash rates for sections with the same characteristics, estimation of critical values for the variables corresponding to the percentile values for average number of crashes, calculation of tables of critical average annual daily traffic values, and generation of a list of TWLTL locations with critical safety concerns. Results presented in the paper have been used in real applications. Copyright © 2012 John Wiley & Sons, Ltd.     

Disaggregate safety evaluation for signalized intersections and evaluation tool

This study was to evaluate traffic safety of four‐legged signalized intersections and to develop a spreadsheet tool for identifying high‐risk intersections taking into consideration vehicle movements, left‐turn signal phase types, and times of day. The study used data from Virginia and employed count data models and the empirical Bayes (EB) method for safety evaluation of such intersections. It was found that crash pattern defined by vehicle movements involved in a crash and time of day are important factors for intersection crash analysis. Especially for a safety performance function (SPF), a model specification (Poisson or NB), inclusion of left‐turn signal types, type of traffic flow variables, variable functional forms, and/or magnitudes of coefficients turned out to be different across times of day and crash patterns. The spreadsheet application tool was developed incorporating the developed SPFs and the EB method. As long as Synchro files for signal plans and crash database are maintained, no additional field data collection efforts are required. Adjusting the developed SPFs and the spreadsheet for recent traffic and safety conditions can be done by applying the calibration methods employed in the SafetyAnalyst software and the Highway Safety Manual. Implementing the developed tool equipped with streamlining data entry would greatly improve accuracy and efficiency of safety evaluation of four‐legged signalized intersections in localities and highway agencies that cannot operate the SafetyAnalyst. Copyright © 2010 John Wiley & Sons, Ltd.     

The joint analysis of injury severity of drivers in two-vehicle crashes accommodating seat belt use endogeneity

The current study contributes to the existing injury severity modeling literature by developing a multivariate probit model of injury severity and seat belt use decisions of both drivers involved in two-vehicle crashes. The modeling approach enables the joint modeling of the injury severity of multiple individuals involved in a crash, while also recognizing the endogeneity of seat belt use in predicting injury severity levels as well as accommodating unobserved heterogeneity in the effects of variables. The proposed model is applied to analyze the injury severity of drivers involved in two-vehicle road crashes in Denmark.The empirical analysis provides strong support for the notion that people offset the restraint benefits of seat belt use by driving more aggressively. Also, men and those individuals driving heavy vehicles have a lower injury risk than women and those driving lighter vehicles, respectively. At the same time, men and individuals driving heavy vehicles pose more of a danger to other drivers on the roadway when involved in a crash. Other important determinants of injury severity include speed limit on roadways where crash occurs, the presence (or absence) of center dividers (median barriers), and whether the crash involves a head-on collision. These and other results are discussed, along with implications for countermeasures to reduce injury severities in crashes. The analysis also underscores the importance of considering injury severity at a crash level, while accommodating seat belt endogeneity effects and unobserved heterogeneity effects.     

Traffic models for goods,service and business movements to manufacturing establishments

This paper is concerned with a problem area of increasing interest, namely the traffic characteristics of specific generators located on non‐residential land use. It concentrates on the problem of modelling traffic generation for goods, service and business movements at the manufacturing establishment. Firstly the contributions of earlier researchers are reviewed in the context of factors which have been considered important. Then a further study conducted by the author in Sheffield and in Manchester, England, is described. This study considers more rigorously than hitherto the questions of similarities in trip rates over different manufacturing activities and geographical areas, and of the choice of explanatory variables. Comparison of regressions is used to formalize indications of similarities in trip rates, and analysis of variance for choosing the most satisfactory explanatory variables. The study provides evidence of similarity in trip rates over seven different manufacturing activities and two geographical areas. It also shows distinctions between different non‐work traffic types in the extent of similarities in trip rates over different manufacturing activities and in explanatory variables for traffic generation relationships. Regarding the latter it is concluded that it is not possible to model accurately such traffic movements on the basis of explanatory variables which are normally available. Instead trip rates should be expressed in simple mathematical terms and should be based on large data populations. The study also makes contributions on peak flow factors for the actual design of facilities of manufacturing establishments. These findings and others are discussed in the context of the work of earlier researchers. Finally conclusions are drawn and guidance offered for future work into this problem.     

Household travel attributes transferability analysis: application of a hierarchical rule based approach

This paper presents a system of hierarchical rule-based models of trip generation and modal split. Travel attributes, like trip counts for different transportation modes and commute distance, are among the modeled variables. The proposed framework could be considered as an alternative for several modules of the traditional travel demand modeling approach, while providing travel attributes at the highly disaggregate level that can be also used in activity-based micro-simulation modeling systems. Nonetheless, the modeling framework of this study is not considered as a substitute for activity-based models. The explanatory variables set ranges from socio-economic and demographic attributes of the household to the built environment characteristics of the household residential location. Another important contribution of the study is a framework in which travel attributes are modeled in conjunction with each other and the interdependencies among them are postulated through a hierarchical system of models. All the models are developed using rule-based decision tree method. Moreover, the models developed in this study present a useful improvement in increasing the practicality and accuracy of the rule-based travel data simulation models.     

Are we ready to embrace connected and self-driving vehicles? A case study of Texans

While connected, highly automated, and autonomous vehicles (CAVs) will eventually hit the roads, their success and market penetration rates depend largely on public opinions regarding benefits, concerns, and adoption of these technologies. Additionally, the introduction of these technologies is accompanied by uncertainties in their effects on the carsharing market and land use patterns, and raises the need for tolling policies to appease the travel demand induced due to the increased convenience. To these ends, this study surveyed 1088 respondents across Texas to understand their opinions about smart vehicle technologies and related decisions. The key summary statistics indicate that Texans are willing to pay (WTP) $2910, $4607, $7589, and $127 for Level 2, Level 3, and Level 4 automation and connectivity, respectively, on average. Moreover, affordability and equipment failure are Texans’ top two concerns regarding AVs. This study also estimates interval regression and ordered probit models to understand the multivariate correlation between explanatory variables, such as demographics, built-environment attributes, travel patterns, and crash histories, and response variables, including willingness to pay for CAV technologies, adoption rates of shared AVs at different pricing points, home location shift decisions, adoption timing of automation technologies, and opinions about various tolling policies. The practically significant relationships indicate that more experienced licensed drivers and older people associate lower WTP values with all new vehicle technologies. Such parameter estimates help not only in forecasting long-term adoption of CAV technologies, but also help transportation planners in understanding the characteristics of regions with high or low future-year CAV adoption levels, and subsequently, develop smart strategies in respective regions.     

Real time traffic information and its impacts on route switching behavior of expressway drivers

Non‐quantifiable factors (e.g. perceived, attitudinal and preferential factors) have not been investigated fully in past transportation studies, which has raised questions on the predictive capabilities of the models. In this study, Structure Integration Models, with one of their sub‐models, Measurement Equation, are combined with latent variables, which are integrated with another sub‐model, Structural Equation. The estimated latent variables are used as explanatory variables in decision models. As a result, the explanatory and predictive capabilities of the models are enhanced. The models can then be used to describe the various behaviors of travelers of different types of transportation systems in a more accurate way. In this study, the Structure Integration Model was applied to study the impacts of real‐time traffic information on the route‐switching behavior of road users on the Sun Yat‐Sen expressway, Taiwan. At present, the real‐time traffic information provided on this expressway includes radio traffic reports and changeable message signs. The results of this study can facilitate the provision of traffic information on highways.     

Investigating the nature and impact of reporting bias in road crash data

This paper investigates the nature, and impact of the reporting bias associated with the police-reported crash data on inferences made using this data. In doing so, we merge a detailed emergency room data and police-reported crash data for a specific region in Denmark. To disentangle potentially common observable and unobservable factors that affect drivers’ injury severity risk and their crash reporting behavior, we formulate a bivariate ordered-response probit model of injury severity risk and crash reporting propensity. To empirically identify the reporting bias in this joint model, we exploit an exogenous police reform that particularly affects some specific municipalities of the region under consideration. The empirical analysis reveals substantial reporting bias in the commonly used police-reported road crash data. This non-random sample selection associated with the police-reported crash data leads to biased estimates on the effect of some of the explanatory variables in injury severity analysis. For instance, estimates based on the police-reported crash data substantially underestimate the effectiveness of seat belt use in reducing drivers’ injury severity risk.     

A topological route choice model for metro

This article presents a route choice model for public transit networks that incorporates variables related to network topology, complementing those found in traditional models based on service levels (travel time, cost, transfers, etc.) and users’ socioeconomic and demographic characteristics (income level, trip purpose, etc.). The topological variables represent concepts such as the directness of the chosen route and user knowledge of the network. For both of these factors, the necessary data is endogenous to the modelling process and can be quantified without the need for information-gathering beyond what is normally required for building route choice models. Other novel variables in the proposed formulation capture notions of user comfort such as vehicle occupancy rates and certain physical characteristics of network stations. We conclude that these new variables significantly improve the explanatory and predictive ability of existing route choice specifications.     

On accommodating spatial dependence in bicycle and pedestrian injury counts by severity level

This paper proposes a new spatial multivariate count model to jointly analyze the traffic crash-related counts of pedestrians and bicyclists by injury severity. The modeling framework is applied to predict injury counts at a Census tract level, based on crash data from Manhattan, New York. The results highlight the need to use a multivariate modeling system for the analysis of injury counts by road-user type and injury severity level, while also accommodating spatial dependence effects in injury counts.     

Specification and estimation of a spatially and temporally autocorrelated seemingly unrelated regression model: application to crash rates in China

In transportation studies, variables of interest are often influenced by similar factors and have correlated latent terms (errors). In such cases, a seemingly unrelated regression (SUR) model is normally used. However, most studies ignore the potential temporal and spatial autocorrelations across observations, which may lead to inaccurate conclusions. In contrast, the SUR model proposed in this study also considers these correlations, making the model more behaviorally convincing and applicable to circumstances where a three-dimensional correlation exists, across time, space, and equations. An example of crash rates in Chinese cities is used. The results show that incorporation of spatial and temporal effects significantly improves the model. Moreover, investment in transportation infrastructure is estimated to have statistically significant effects on reducing severe crash rates, but with an elasticity of only −0.078. It is also observed that, while vehicle ownership is associated with higher per capita crash rates, elasticities for severe and non-severe crashes are just 0.13 and 0.18, respectively; much lower than one. The techniques illustrated in this study should contribute to future studies requiring multiple equations in the presence of temporal and spatial effects.

Investigating effects of asphalt pavement conditions on traffic accidents in Tennessee based on the pavement management system (PMS)

Pavement maintenance is essential for ensuring good riding quality and avoiding traffic congestion, air pollution, and accidents. Improving road safety is one of the most important objectives for pavement management systems. This study utilized the Tennessee Pavement Management System (PMS) and Accident History Database (AHD) to investigate the relationship between accident frequency and pavement distress variables. Focusing on four urban interstates with asphalt pavements, divided median types, and 55 mph speed limits, 21 Negative Binomial Regression models were developed for predicting various types of traffic accident frequencies based on different pavement condition variables, including rut depth (RD), International Roughness Index (IRI), and Present Serviceability Index (PSI). The modeling results indicated that the RD models did not perform well, except for predicting accidents at night and accidents under rain weather conditions; whereas, IRI and PSI were always significant prediction variables in all types of accident models. Comparing the models goodness‐of‐fit results, it was found that the PSI models had a better performance in crash frequency prediction than the RD models and IRI models. This study suggests that the PSI accident prediction models should be considered as a comprehensive approach to integrate the highway safety factors into the pavement management system. Copyright © 2010 John Wiley & Sons, Ltd.     

A new logit‐artificial neural network ensemble for mode choice modeling: a case study for border transport

Logit model is one of the statistical techniques commonly used for mode choice modeling, while artificial neural network (ANN) is a very popular type of artificial intelligence technique used for mode choice modeling. Ensemble learning has evolved to be very effective approach to enhance the performance for many applications through integration of different models. In spite of this advantage, the use of ANN‐based ensembles in mode choice modeling is under explored. The focus of this study is to investigate the use of aforementioned techniques for different number of transportation modes and predictor variables. This study proposes a logit‐ANN ensemble for mode choice modeling and investigates its efficiency in different situations. Travel between Khobar‐Dammam metropolitan area of Saudi Arabia and Kingdom of Bahrain is selected for mode choice modeling. The travel on this route can be performed mainly by air travel or private vehicle through King Fahd causeway. The results show that the proposed ensemble gives consistently better accuracies than single models for multinomial choice problems irrespective of number of input variables. Copyright © 2015 John Wiley & Sons, Ltd.