Analysis of real-time crash risk for expressway ramps using traffic, geometric, trip generation, and socio-demographic predictors

Analysis of harsh braking and harsh acceleration occurrence via explainable imbalanced machine learning using high-resolution smartphone telematics and traffic data

Harsh driving events such as harsh brakings (HBs) and harsh accelerations (HAs) are promising Surrogate Safety Measures, already extensively utilised in road safety research. However, their occurrence relative to normal driving conditions has not been the explicit target of research, as they are typically used as inputs for crash prediction. The present study addresses this research gap by investigating factors influencing HB and HA occurrence using real-time naturalistic driving telematics data recorded from smartphones, traffic data and road geometry & network characteristics data. These multisource data were matched in order to capture the specific circumstances under which HBs and HAs occur. The utilized telematics dataset included trips from 314 anonymous drivers in an urban arterial of Athens, Greece. Subsequently, Synthetic Minority Oversampling TEchnique (SMOTE) was applied due to class imbalance and then binary classification was conducted to detect factors leading to HB and HA occurrence. Imbalanced Machine Learning (ML) XGBoost algorithms predicted over 75% of HBs and over 84% of HAs for the test dataset, indicating suitability for real-time monitoring. The algorithms were also augmented with SHapley Additive exPlanation (SHAP) values, aiming to increase outcome explainability. Results reveal strong nonlinear effects on harsh event occurrence, with individual speed and traffic flow parameters showing the highest influence, followed by exposure parameters such as segment length and pass count. Network characteristics such as number of lanes, and speed limit had limited influence on HA and HB occurrence, as did behaviors such as mobile phone engagement and speeding.

Real-time crash prediction on express managed lanes of Interstate highway with anomaly detection learning

To facilitate efficient transportation, I-4 Express is constructed separately from general use lanes in metropolitan area to improve mobility and reduce congestion. As this new infrastructure would undoubtedly change the traffic network, there is a need for more understanding of its potential safety impact. Unfortunately, many advanced real-time crash prediction models encounter an important challenge in their applicability due to their demand for a substantial volume of data for direct modeling. To tackle this challenge, we proposed a simple yet effective approach - anomaly detection learning, which formulates model as an anomaly detection problem, solves it through normality feature recognition, and predicts crashes by identifying deviations from the normal state. The proposed approach demonstrates significant improvement in the Area Under the Curve (AUC), sensitivity, and False Alarm Rate (FAR). When juxtaposed with the prevalent direct classification paradigm, our proposed anomaly detection learning (ADL) consistently outperforms in AUC (with an increase of up to 45%), sensitivity (experiencing up to a 45% increase), and FAR (reducing by up to 0.53). The most performance gain is attained through the combination of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) in an ensemble, resulting in a 0.78 AUC, 0.79 sensitivity, and a 0.22 false alarm rate. Furthermore, we analyzed model features with a game-theoretic approach illustrating the most correlated features for accurate prediction, revealing the attention of advanced convolution neural networks to occupancy features. This provided crucial insights into improving crash precaution, the findings from which not only benefit private stakeholders but also extend a promising opportunity for governmental intervention on the express lane. This work could promote express lane with more efficient resource allocation, real-time traffic management optimization, and high-risk area prioritization.

Who might encounter hard-braking while speeding? Analysis for regular speeders using low-frequency taxi trajectories on arterial roads and explainable AI

Regular speeders are those who commit speeding recidivism during a period. Among their speeding behaviors, some occurring in specific scenarios may cause more hazards to road users. Therefore, there is a need to evaluate the driving risks if the regular speeders have different speeding propensities. This study considers speeding-related hard-braking events (SHEs) as a safety surrogate measure and recognizes the regular speeders who encounter at least one SHEs during the study period as risky individuals. To identify speeding behaviors and hard-braking events from low-frequency GPS trajectories, we compare the average travel speed between pairwise adjacent GPS points to the posted speed limit and examine the speed curve and the corresponding travel distance between these GPS points, respectively. Thereafter, a logistic model, XGBoost, and three 1D Convolutional Neural Networks (CNNs) including AlexNet CNN, Mini-AlexNet CNN, and Simple CNN are respectively developed to recognize the regular speeders who encountered SHEs based on their speeding propensities. The proposed Mini-AlexNet CNN achieves a global F1-score of 91% and recall of 90% on the testing data, which are superior to other models. Further, the study uses the Shapley Additive exPlanation (SHAP) framework to visually interpret the contribution of speeding propensities on SHE likelihood. It is found that speeding by 50% or greater for no more than 285 m is the most dangerous kind among all the speeding behaviors. Speeding on roads without bicycle lanes or on roads with roadside parking and excessive accesses increases the probability of encountering SHEs. Based on the analyses, we put forward tailored recommendations that aim to restrict hazard-related speeding behaviors rather than speeding behaviors of all kinds.

Prediction and analysis of likelihood of freeway crash occurrence considering risky driving behavior

The prediction of the likelihood of vehicle crashes constitutes an indispensable component of freeway safety management. Due to data collection limitations, studies have used mainly traffic flow-related variables to develop freeway crash prediction models but rarely have considered the effect of risky driving behavior on the likelihood of crashes. This study employed navigation software to collect driving behavior data and integrated multi-source data that include vehicle speed, traffic volume, and congestion index values. The study also employed the 'synthesizing minority oversampling technique and edited nearest neighbor' (SMOTE + ENN) coupled method for data balance processing. Three freeway crash likelihood prediction models were built based on the binomial logit, eXtreme Gradient Boosting (XGBoost), and support vector machine algorithms, respectively. The Shapley additive explanation (SHAP) algorithm was utilized to explore the effect of each feature variable on the likelihood of crashes. The results show that the prediction accuracy of the XGBoost model is the best of the three compared models. Under the optimal control-to-case ratio (1:1), the prediction accuracy of the XGBoost model reached 0.96 in this study, and the recall rate, specificity, and area-under-the-curve values were 0.86, 0.96, and 0.907, respectively. Comparative test results demonstrate that ranking risky driving behavior into three levels of intensity can effectively enhance the predictive accuracy of the XGBoost model. Moreover, the XGBoost model with its ten-minute time step outperformed the XGBoost model with its five-minute time step in terms of prediction accuracy. The results of the SHAP-based analysis show that the likelihood of highway crashes is high when the traffic congestion level is high and the distribution of the vehicle speed in the upstream roadway section is significant. Also, both sharp acceleration and sharp deceleration lead to greater likelihood of crashes. This paper aims to provide an effective framework for predicting and interpreting the likelihood of freeway crashes, thereby providing guidance for crash prevention, driver training, and the development of traffic regulations.

An integrated intelligent intersection control system for preventing rear-end and angled crashes: System design and deployment results

In view of the dynamic all-red extension (DARE) system's effectiveness in preventing angled crashes (Park et al., 2018), this study has further enhanced its function to contend with rear-end collisions with dynamic green extension (DGE). With such a function, the enhanced Integrated Intelligent Intersection control system (III-CS) is capable of dynamically terminating the green at the interval of the lowest rear-end collision risk, so as to prevent undesirable "max-out" under actuated signal control which often traps some vehicles in the dilemma zone during high-volume traffic conditions. To ensure its effectiveness in practice, the proposed III-CS has been designed with the following new features: (i) executing the DGE within a customized time window of the green phase to ensure the signal's effective coordination with its neighboring intersections; (ii) adopting the comparison-based heuristic for the DGE's real-time risk prediction so as to circumvent the computing and communications delays. The results of two after-deployment assessments show that the system's DARE has perfectly detected all red-light runners; 66.7 percent of the decisions by the DGE module were observed to achieve the control objective during the first field assessment. The DGE's performance in making optimal decisions has improved over time and reached the level of 81.3% in the second field evaluation. Other measures of effectiveness, such as the number of vehicles trapped in the dilemma zone and the average deceleration rate of the driving populations approaching the target intersection, have also evolved to the anticipated trend after the deployment.

Traffic conflict assessment using macroscopic traffic flow variables: A novel framework for real-time applications

The present study develops a comprehensive traffic conflict assessment framework using macroscopic traffic state variables. To this end, vehicular trajectories extracted for a midblock section of a ten-lane divided Western Urban Expressway in India are used. A macroscopic indicator termed "time spent in conflict (TSC)" is adopted to evaluate traffic conflicts. The proportion of Stopping distance (PSD) is adopted as a suitable traffic conflict indicator. Vehicle-to-vehicle interactions in a traffic stream are two-dimensional, highlighting that the vehicles interact simultaneously in lateral and longitudinal dimensions. Therefore, a two-dimensional framework based on the influence zone of the subject vehicle is proposed and employed to evaluate TSCs. The TSCs are modeled as a function of macroscopic traffic flow variables, namely, traffic density, speed, the standard deviation in speed, and traffic composition, under a two-step modeling framework. In the first step, the TSCs are modeled using a grouped random parameter Tobit (GRP-Tobit) model. In the second step, data-driven machine learning models are employed to model TSCs. The results revealed that intermediately congested traffic flow conditions are critical for traffic safety. Furthermore, macroscopic traffic variables positively influence the value of TSC, highlighting that the TSC increases with an increase in the value of any independent variable. Among different machine learning models, the random forest (RF) model was observed as the best-fitted model to predict TSC based on macroscopic traffic variables. The developed machine learning model facilitates traffic safety monitoring in real-time.

Understanding the impact of vehicle dynamics, geometric and non-geometric roadway attributes on surrogate safety measure using connected vehicle data

Traditional safety research mostly relies on accident data to analyze the precedents to a crash. Alternatively, surrogate safety measures have the potential to proactively evaluate safety events. The era of connected vehicles and smart sensing has brought about tremendous innovations in safety research. GPS data from such vehicles form a useful case of big data analytics where surrogate safety measures have largely been unexplored. In this paper, we propose time to collision estimation from connected vehicle GPS data. The vehicle dynamics such as speed, acceleration, yaw rate, etc. are then coupled with geometric and non-geometric roadway attributes to understand the contributing factors for a traffic conflict. The dataset contains 2,568,421 GPS points from 14,753 unique journeys. 1:4 ratio of conflict to non-conflict events was used to select 15,258 samples with 28 independent vehicle dynamics, geometric, and non-geometric variables. Binary logit model was used to investigate the relationship of these variables with conflicts. Model results showed that out of 28 independent variables, 6 independent variables and 7 interaction variables were found significant. The results showed some interesting and unique relations of these variables with conflicts. Based on these significant variables, k-means clustering was performed to understand the threshold for the significant values for which the number of conflicts is significantly increased. Results from k-means clustering and two sample binomial proportion t-tests revealed that when absolute acceleration crossed 0.8 m/s², conflict probability increased by 8 percentage points.​ Moreover, when the yaw rate crossed 8 degrees/s, the conflict probability doubled. Besides, vehicles traveling at more than 140% of the recommended speed limit increased conflict probability by 7 percentage points.

Developing a new real-time traffic safety management framework for urban expressways utilizing reinforcement learning tree

One of the main objectives of an urban traffic control system is to reduce the crash frequency and the loss caused by these crashes on urban expressways. Real-time crash risk prediction (RTCRP) is an essential technique to identify crash precursors so as to take proactive measures to smooth traffic fluctuations. In addition, automatic incident detection (AID) is another important approach to timely detect an incident so as to design countermeasures that reduce any negative impacts on traffic dynamics. With the introduction of disruptive technologies in transport, highly disaggregated large datasets have started to emerge for modelling while existing modelling techniques utilized in RTCRP and AID may not be able to accurately predict traffic crashes in real-time. Therefore, this paper proposes a state-of-the-art reinforcement learning tree (RLT) approach to develop RTCRP model and automatic crash detection (ACD) model similar to AID, and further utilizes it to build a real-time traffic safety management framework for urban expressways with the input of online traffic data streaming. Recorded traffic flow data and historical crash data were extracted and integrated to develop and implement both RTCRP models and ACD models. The prediction results were compared with the frequently used logistic regression (LR), support vector machine (SVM) and deep neural network (DNN) and a sensitivity analysis for variable effects was conducted. The results confirm that RLT outperforms LR, SVM and DNN in developing RTCRP and ACD models. At the cost of 10% false-alarm rate, about 96% of the crashes were predicted or detected correctly by the proposed framework. The results also indicate that: i) collecting more data is helpful to improve the predictive performance and approximatively a minimum sample size of 20 observations per variable is reasonable for training RLT models; and ii) obtaining more factors is beneficial to improve the predictive performance. With the RLT approach, it was demonstrated that selected important variables also have the capability to provide reasonable predictive performance.

Plasma Lipid-based Machine Learning Models Provides a Potential Diagnostic Tool for Colorectal Cancer Patients

Colorectal cancer is the second leading cause of cancer-related death across the world. So far, screening methods for colorectal cancer are limited to blood test, imaging test, and digital rectal examination, that are either invasive or ineffective. So, this study aims to explore novel, more convenient and effective diagnostic methods for colorectal cancer. First, the experiment cohort was randomly split to train set and test set, and LC-MS-based plasma lipidomics was applied to identify lipid features in colorectal cancer. Second, univariate and multivariate analyses were performed to screen for significantly differentially expressed lipids. Third, single-lipid-based ROC analysis and multiple-lipid-based machine learning modelling were conducted to assess differential lipids' diagnostic performance. Lastly, survival analyses were used to evaluate lipids' prognostic values. In total, 41 differential lipids were screened out, 10 were upregulated and 31 were downregulated in CRC. Only CerP(d15:0_22:0+O) showed fine predictive accuracy in single-lipid-base ROC analysis. Among the four machine learning models, SVM showed best predictive performance with accuracy (in predicting test set) of 1.0000 (95%CI: 0.8806, 1.0000), that can be reached by modelling with only 14 lipids. Four lipids had significant prognostic values, that were TG(11:0_18:0_18:0) (HR: 0.34), TG(18:0_18:0_18:1) (HR: 0.34), PC(22:1_12:3) (HR: 2.22), LPC(17:0) (HR: 3.16). In conclusion, this study discovered novel lipid features that has potential diagnostic and prognostic values, and showed combination of plasma lipidomics and machine learning modelling could have outstanding diagnostic performance and may serve as a convenient and more accessible way to aid clinical diagnosis of colorectal cancer.

Temporal-spatial evolution analysis of severe traffic violations using three functional forms of models considering the diurnal variation of meteorology

Traffic violations and crashes are inherently associated. Analysis of traffic violation frequency is a prerequisite for improvements in crash prevention and corresponding countermeasures. One of the essential works in the field of traffic violations relates to the exploration of the correlations between a certain violation type (e.g., speeding or safety belt use) and its causal factors (e.g., demographics and road types). Till now, the effects of spatiotemporal and meteorological factors on severe traffic violations, a general term for dangerous driving behaviors, have not been fully considered. Using the dataset consisting of daily severe traffic violations and meteorological conditions during 12 months in Jiangsu Province, China, violation performance functions were developed for three violation types (total violations, driving under the influence, and speeding) based on three models (Poisson regression, zero-inflated Poisson regression, and negative binomial model). The findings indicate that the negative binomial model has a better performance for traffic violation frequency estimation. Additionally, elastic analysis for three violation types relying on the negative binomial model was conducted to present the relationships between the explanatory variables and the expected violation frequency. The effects of spatiotemporal factors have revealed that the violation situations are significantly different in varying cities and the frequency of drunk driving shows a significant time instability. It is also found that rainy days will generate a decrease in the possibility of violation occurrence. With regard to temperature, a significant negative effect is found and the decrease in temperature will bring about an increase in violation frequency. Besides, traffic violation frequency is significantly increased during holidays with comfortable weather conditions. The conclusion of this study can provide insightful suggestions for the department of traffic enforcement to adjust the patrol plans according to the specified periods (weeks, months, or holidays) and weather conditions. Special rectification actions and targeted educational activities are also advised to be put forward simultaneously.

Risk Identification and Conflict Prediction from Videos Based on TTC-ML of a Multi-Lane Weaving Area

Crash risk identification and prediction are expected to play an important role in traffic accident prevention. However, most of the existing studies focus only on highways, not on multi-lane weaving areas. In this paper, a potential collision risk identification and conflict prediction model based on extending Time-to-Collision-Machine Learning (TTC-ML) for multi-lane weaving zone was proposed. The model can accurately learn various features, such as vehicle operation characteristics, risk and conflict distributions, and physical zoning characteristics in the weaving area. Specifically, TTC was used to capture the collision risk severity, and ML extracted vehicle trajectory features. After normalizing and dimensionality reduction of the vehicle trajectory dataset, Naive Bayes, Logistic Regression, and Gradient Boosting Decision Tree (GBDT) models were selected for traffic conflict prediction, and the experiments showed that the GBDT model outperforms two remaining models in terms of prediction accuracy, precision, false-positive rate (FPR) and Area Under Curve (AUC). The research findings of this paper help traffic management departments develop and optimize traffic control schemes, which can be applied to Intelligent Vehicle Infrastructure Cooperative Systems (IVICS) dynamic warning.

Transfer learning-based highway crash risk evaluation considering manifold characteristics of traffic flow

The main objective of this study is to evaluate highway crash risk and improve the spatial and temporal transferability of crash risk models. The predictive performance is affected by the difficulty of existing models to quantify crash risk from historical traffic flow data at different locations and times and may not fully capture the complex nonlinear relationships between high-dimensional factors in traffic flow states. Oregon US26W freeway data from 2016 and 2017 and I5N freeway data from 2017 were used. Raw detector data collected from two consecutive detector stations upstream-downstream detector stations were converted into 30 traffic variables. The averages, standard deviations, and coefficients of variation were obtained by aggregating traffic values using each lane. Candidate variables for traffic flow were extracted, and the importance of each variable was calculated using LightGBM, which reveals that variable differences between lanes contributed more. The manifold distance was then applied to quantify the crash risk and classify traffic crashes or not. When the manifold distance is 0.4, it could effectively distinguish traffic crashes. TransferBoost was further employed to build a crash risk model. Modeling using 2016 and 2017 data from the US26W freeway revealed a significant decrease in AUC and a gradual decrease in the model's sensitivity. However, the crash risk prediction performance of TransferBoost improved by 5.2% when modeling using 2017 data from US26W and I5N freeways. The results show that the model developed for one time period cannot be directly used to predict crash risk for another period on the same freeway. However, the model developed for one highway cannot be directly used to predict the crash risk of another highway either, maintaining some transferability at low false alarm rates. TransferBoost provides a fresh perspective on the transferability of the model. The findings of this study could facilitate more accurate proactive safety management and improvement countermeasure development.

A multinomial logit model: Safety risk analysis of interchange area based on aggregate driving behavior data

INTRODUCTION

Urban expressway interchanges have become accident-prone sites owing to the accelerated increase in motor-vehicle ownership. This study explored the impact of factors, including day of the week, time of day, congestion level, traffic control devices, and road conditions, on road safety risk levels in the interchange area of an urban expressway based on aggregate driving behavior data.

METHOD

A large amount of aggregate driving behavior data were obtained from AutoNavi navigation software. The database was built by matching various types of data and observing their characteristics. Day of the week, time of day, congestion level, road conditions (number of lanes, traffic disturbance, and traffic control devices [the type of advance guide sign system, number of warning signs, and the complexity of the diagrammatic guide sign]) were identified as the explanatory variables. The traffic order index (TOI), based on driving behavior and speed variation, was used to evaluate the road safety risk levels, including risky roads, general roads, and safe roads, which served as the response variables. The multinomial logit model (MNL) was developed to explore the impact of various factors, including traffic control devices and road conditions, on road safety risk levels.

RESULTS

The results showed that the factors that significantly influence risky roads include day of the week, number of lanes, congestion level (slow moving), traffic disturbance (with the merge or diverge within 500 m), type of advance guide sign system (three-level advance guide sign system), and complexity of diagrammatic guide signs (low or medium complexity). Practical Applications: This study could offer plausible suggestions for traffic management departments for the rehabilitation of road conditions and traffic control devices in urban expressway interchange areas.

A resampling approach to disaggregate analysis of bus-involved crashes using panel data with excessive zeros

Public bus constitutes more than 70% of the overall road-based public transport patronage in Hong Kong, and its crash involvement rate has been the highest among all public transport modes. Though previous studies had identified explanatory factors that affect the crash risk of buses, use of considerably imbalanced crash data with excessive zero observations could lead to inaccurate parameter estimation. This study aims to resolve the excess zero problem of disaggregate analysis of bus-involved crashes based on synthetic data using a Synthetic Minority Over-Sampling Technique for panel data (SMOTE-P). Dataset comprising crash, traffic, and road inventory data of 88 road segments in Hong Kong during the period from 2014 to 2017 is used. To assess the data balancing performance, other common data generation approaches such as Random Under-sampling of the Majority Class (RUMC) technique, Cluster-Based Under-Sampling (CBUS), and mixed resampling, are also considered. Random effect Poisson (REP) models based on synthetic data and random effect zero-inflated Poisson (REZIP) model based on original data are estimated. Results indicate that REP model based on synthetic data using SMOTE-P outperforms REZIP model based on original data and REP models based on synthetic data using RUMC, CBUS and mixed approaches, in terms of statistical fit, prediction error, and explanatory factors identified. Results of model estimation based on SMOTE-P suggest that factors including morning peak, evening peak, hourly traffic flow, average lane width, road length, bus stop density, percentage of bus in the traffic stream, and presence of bus priority lane all affect the bus-involved crash frequency. More importantly, this study provides a feasible solution for disaggregate crash analysis with imbalanced panel data.

A conflict-based approach for real-time road safety analysis: Comparative evaluation with crash-based models

An innovative approach for real-time road safety analysis is presented in this work. Unlike traditional real-time crash prediction models (RTCPMs), in which crash data are used in the training phase, a real-time conflict prediction model (RTConfPM) is proposed. This model can be trained using surrogate measures of safety, and can therefore be applied even in situations in which highly spatial/temporal-accurate crash data are unavailable or unreliable. The application of an RTConfPM consists of using a set of input variables recorded during a given time interval, to predict whether there will be an increased risk of unsafe situations in the following interval. This paper presents an RTConfPM to predict rear-end crashes, using time-to-collision values recorded with radar sensors on multiple motorway cross-sections to define unsafe situations, and traffic conditions recorded on the same sections as input to the model. The RTConfPM is compared to a traditional RTCPM, trained with a dataset of crashes located on the same motorway, and using the same traffic data as input. In both approaches, variable selection is performed with Pearson's correlation test and random forest; synthetic minority oversampling technique (SMOTE) is used to balance the classes in the training dataset, support vector machine (SVM) is used as classifier, and Monte Carlo cross-validation is adopted for robustness. The two approaches are evaluated considering accuracy, recall, specificity/false alarm rate, and area under the curve (AUC). As shown by the results of this paper, the conflict-based approach appears promising, and is able to predict the occurrence of unsafe situations within 5 min with more than 93% accuracy, recall and specificity, significantly outperforming the RTCPM.

Investigating the predictability of crashes on different freeway segments using the real-time crash risk models

Improvement of the prediction efficiency of crash risks has attracted the attention of numerous studies. Nevertheless, one of the most important factors, crash precursors, were neglected. This study mainly focuses on identifying optimal crash precursors for different freeway section types, as well as providing a threshold selection method for real-time crash risk models. Freeway sections are divided into four types, i.e. basic sections, weaving areas, merging areas, and diverging areas. Bayesian logistic regression (BLR) models were established for each type of segment, and significant factors were distinguished. A threshold selection method was proposed based on cost-benefit theory, and the threshold is determined as the value when the number of proactive safety interventions to prevent a crash is 5000 in this study. BLR models with one, two and three optimal variables were developed. Then the sensitivity and false alarm rate of the models were obtained and compared. Comparison results show that the minimum amount of parameters which can achieve the ideal prediction effectiveness is two. In this situation, 25 %, 50 %, 20 % and 20 % of the crashes occurring at basic sections, weaving areas, merging areas and diverging areas can be accurately predicted respectively. Downstream average speed was recommended as the best crash precursor variable for all the segment types. Support Vector Machine and Random Forest were applied to confirm the conclusion. The conclusion of this paper has the possibility to help reduce crash risk to a relatively economical level in practical applications.

Predicting unsafe driving risk among commercial truck drivers using machine learning: Lessons learned from the surveillance of 20 million driving miles

The emergence of sensor-based Internet of Things (IoT) monitoring technologies have paved the way for conducting large-scale naturalistic driving studies, where continuous kinematic driver-based data are generated, capturing crash/near-crash safety critical events (SCEs) and their precursors. However, it is unknown whether the SCEs risk can be predicted to inform driver decisions in the medium term (e.g., hours ahead) since the literature has focused on SCE predictions either for a given road segment or for automated breaking applications, i.e., immediately before the event. In this paper, we examine the SCE data generated from 20+ million miles-driven by 496 commercial truck drivers to address three main questions. First, whether SCEs can be predicted using disparate driving-related data sources. Second, if so, what the relative importance of the different predictors examined is. Third, whether the prediction models can be generalized to new drivers and future time periods. We show that SCEs can be predicted 30 min in advance, using machine learning techniques and dependent variables capturing the driver's characteristics, weather conditions, and day/time categories, where an area under the curve (AUC) up to 76% can be achieved. Moreover, the predictive performance remains relatively stable when tested on new (i.e., not in the training set) drivers and a future two-month time period. Our results can inform dispatching and routing applications, and lead to the development of technological interventions to improve driver safety.

A study of freeway crash risk prediction and interpretation based on risky driving behavior and traffic flow data

The prediction of traffic crashes is an essential topic in traffic safety research. Most of the previous studies conducted experiments on real-time crash prediction of expressways or freeways, based on traffic flow data. However, the influence of risky driving behavior on traffic crash risk prediction has rarely been considered. Thus, a traffic crash risk prediction model based on risky driving behavior and traffic flow has been developed. The data employed in this research were captured using the in-vehicle AutoNavigator software. A random forest to select variables with strong impacts on crashes and the synthetic minority oversampling technique (SMOTE) to adjust the imbalanced dataset were included in the research. A logistic regression model was developed to predict the risk of traffic crash and to interpret its relationship with traffic flow and risky driving behavior characteristics. This model accurately predicted 84.48% of the crashes, while its false alarm rate remained as low as 9.75%, which indicated that this traffic crash risk prediction model had high accuracy. By analyzing the relationship between traffic flow, risky driving behavior, and crashes through partial dependency plots (PDPs), the impact of traffic flow and risky driving behavior variables on certain traffic crashes in the prediction model were determined. Through this study, the data of traffic flow and risky driving behavior could be used to assess the traffic crash risk on freeways and lay a foundation for traffic safety management.

Speed violation analysis of heavy vehicles on highways using spatial analysis and machine learning algorithms

With the development of technology in the world, vehicles that reach high speeds are produced. In addition, with the increase of road width and quality, faster and more comfortable transportation can be provided. These developments also increase the speed violation rates of road vehicles. Drivers who violate speed limits can endanger both their own lives and the lives of others. Speed violations, of especially heavy vehicles, involve much greater risks than that of light vehicles. Heavy vehicles can cause more serious losses of lives and property in accidents, compared to the ones caused by light vehicles, as they can carry much more freight or passengers than light vehicles. In this study, data regarding the speed violations committed by heavy vehicles in Turkey, were used. Speed violations were divided into 10 classes according to the intensity of speed violation rates. After this process, all provinces were classified according to support vector machines (SVM), naive bayes (NB) and k-nearest neighbors (KNN) algorithms. When the accuracy values and error scales of all three algorithms are examined, it has been determined that the algorithm that gives the most accurate results is the NB algorithm. Based on the classification of this algorithm, speed violation density maps of types of heavy vehicles in Turkey were created by using spatial analysis. According to the density maps, the provinces with the highest speed violations were identified. In the results, it was determined that the rate of heavy vehicle speed violation was highest in the cities such as Erzurum, Konya, and Muğla. Later, these cities were examined in terms of heavy vehicle mobility. At the end of this study, measures were proposed to reduce these violations in cities where speeding violations are intense. Material and moral damages can be prevented, to a great extent, with the implementation of recommendations of policymakers which can reduce speed violations.

Crash data augmentation using variational autoencoder

In this paper, we present a data augmentation technique to reproduce crash data. The dataset comprising crash and non-crash events are extremely imbalanced. For instance, the dataset used in this paper consists of only 625 crash events for over 6.5 million non-crash events. Thus, learning algorithms tend to perform poorly on these datasets. We have used variational autoencoder to encode all the events into a latent space. After training, the model could successfully separate crash and non-crash events. To generate data, we sampled from the latent space containing crash data. The generated data was compared with the real data from different statistical aspects. t-Test, Levene-test and Kolmogrove Smirnov test showed that the generated data was statistically similar to the real data. It was also compared to some of the minority oversampling techniques like SMOTE and ADASYN as well as the GAN framework for generating data. Crash prediction models based on Logistic Regression (LR), Support Vector Machine (SVM) and Artificial Neural Network (ANN) were used to compare the generated data from the different oversampling techniques. Overall, variational autoencoder (VAE) showed excellent results compared to the other data augmentation methods. Specificity is improved by 8% and 4% for VAE-LR and VAE-SVM respectively when compared to SMOTE while the sensitivity is improved by 6% and 5% when compared to ADASYN. Moreover, VAE generated data also helps to overcome the overfitting problem in SMOTE and ADASYN since there is flexibility in choosing the decision boundary.

Injury risk assessment based on pre-crash variables: The role of closing velocity and impact eccentricity

Thorough evaluations on injury risk (IR) are fundamental for guiding interventions toward the enhancement of both the road infrastructure and the active/passive safety of vehicles. Well-established estimates are currently based on IR functions modeled on post-crash variables, such as velocity change sustained by the vehicle (ΔV); thence, these analyses do not directly suggest how pre-crash conditions can be modified to allow for IR reduction. Nevertheless, ΔV can be disaggregated into two contributions which enable its apriori calculation, based only on the information available at the impact instant: the Crash Momentum Index (CMI), representing impact eccentricity at collision, and the closing velocity at collision (V_r). By extensively employing the CMI indicator, this work assesses the overall influence of impact eccentricity and closing velocity on the risk for occupants to sustain a serious injury. As CMI synthesizes indications regarding ΔV, its use can be disjointed from the ΔV itself for the derivation of high-quality IR models. This feature distinguishes CMI from the other eccentricity indicators available at the state-of-the-art, allowing for the contribution of eccentricity on IR to be completely isolated. Because of this element of originality, special attention is given to the CMI variable throughout the present work. Based on data extracted from the NASS/CDS database, the influence of the CMI and V_r variables on IR is specifically highlighted and analyzed from several perspectives. The feature ranking algorithm ReliefF, whose use is unprecedented in the accident analysis field, is first employed to assess importance of such impact-related variables in determining the injury outcome: if compared to vehicle-related and occupant-related variables (as category and age, respectively), the higher influence of CMI and V_r is initially highlighted. Secondly, the relevance of CMI and V_r is confirmed by fitting different predictive models: the fitted models which include the CMI predictor perform better than models which neglect the CMI, in terms of classical evaluation metrics. As a whole, considering the high predictive power of the proposed CMI-based models, this work provides valuable tools for the apriori assessment of IR.

A Review of Data Analytic Applications in Road Traffic Safety. Part 1: Descriptive and Predictive Modeling

This part of the review aims to reduce the start-up burden of data collection and descriptive analytics for statistical modeling and route optimization of risk associated with motor vehicles. From a data-driven bibliometric analysis, we show that the literature is divided into two disparate research streams: (a) predictive or explanatory models that attempt to understand and quantify crash risk based on different driving conditions, and (b) optimization techniques that focus on minimizing crash risk through route/path-selection and rest-break scheduling. Translation of research outcomes between these two streams is limited. To overcome this issue, we present publicly available high-quality data sources (different study designs, outcome variables, and predictor variables) and descriptive analytic techniques (data summarization, visualization, and dimension reduction) that can be used to achieve safer-routing and provide code to facilitate data collection/exploration by practitioners/researchers. Then, we review the statistical and machine learning models used for crash risk modeling. We show that (near) real-time crash risk is rarely considered, which might explain why the optimization models (reviewed in Part 2) have not capitalized on the research outcomes from the first stream.

Highway crash detection and risk estimation using deep learning

Crash Detection is essential in providing timely information to traffic management centers and the public to reduce its adverse effects. Prediction of crash risk is vital for avoiding secondary crashes and safeguarding highway traffic. For many years, researchers have explored several techniques for early and precise detection of crashes to aid in traffic incident management. With recent advancements in data collection techniques, abundant real-time traffic data is available for use. Big data infrastructure and machine learning algorithms can utilize this data to provide suitable solutions for the highway traffic safety system. This paper explores the feasibility of using deep learning models to detect crash occurrence and predict crash risk. Volume, Speed and Sensor Occupancy data collected from roadside radar sensors along Interstate 235 in Des Moines, IA is used for this study. This real-world traffic data is used to design feature set for the deep learning models for crash detection and crash risk prediction. The results show that a deep model has better crash detection performance and similar crash prediction performance than state of the art shallow models. Additionally, a sensitivity analysis was conducted for crash risk prediction using data 1-minute, 5-minutes and 10-minutes prior to crash occurrence. It was observed that is hard to predict the crash risk of a traffic condition, 10 min prior to a crash.

A review of spatial approaches in road safety

Spatial analyses of crashes have been adopted in road safety for decades in order to determine how crashes are affected by neighboring locations, how the influence of parameters varies spatially and which locations warrant interventions more urgently. The aim of the present research is to critically review the existing literature on different spatial approaches through which researchers handle the dimension of space in its various aspects in their studies and analyses. Specifically, the use of different areal unit levels in spatial road safety studies is investigated, different modelling approaches are discussed, and the corresponding study design characteristics are summarized in respective tables including traffic, road environment and area parameters and spatial aggregation approaches. Developments in famous issues in spatial analysis such as the boundary problem, the modifiable areal unit problem and spatial proximity structures are also discussed. Studies focusing on spatially analyzing vulnerable road users are reviewed as well. Regarding spatial models, the application, advantages and disadvantages of various functional/econometric approaches, Bayesian models and machine learning methods are discussed. Based on the reviewed studies, present challenges and future research directions are determined.

Calculation of the Risk of Lawsuits over Construction Flaws in Flat Roofs

In order to achieve sustainability objectives in the use of a building, its elements’ operating problems should be minimized. From this premise, a total of 497 cases related to construction flaws in flat roofs were analyzed in this research. A matrix was developed indicating the risk of lawsuits by owners according to the degree of nuisance resulting from the construction flaws studied, their technical importance, and the type of pathological origin. Based on these factors, it is possible to predict a greater or lesser probability of an owner filing a lawsuit—risk factor (F). A wide range was found for this probability, with the largest value being 865 times greater than the smallest value. The value of F was divided into 5 categories to classify the diverse results obtained and determine the number of cases and interrelations ascribed to each category. Additionally, the level of presence of said cases was calculated through the analysis of 3 different demographic aspects, it being noted that a greater purchasing power and a higher concentration of urban population lead to more stringent requirements and, subsequently, to a greater number of lawsuits. With all these results, building quality can be improved while resulting in greater constructive-financial sustainability and in a reduction of the economic resources required of society (fewer lawsuits and associated human resources).