Purpose: The aim of this study was to determine the incidence and pattern of injuries resulting from auto-tricycle crashes among patients in a tertiary referral centre in Ghana. Methods: Data were retrospectively extr...Purpose: The aim of this study was to determine the incidence and pattern of injuries resulting from auto-tricycle crashes among patients in a tertiary referral centre in Ghana. Methods: Data were retrospectively extracted from hospital records of patients who got involved in auto-tricycle crashes and presented to the Accident and Emergency Centre of the Komfo Anokye Teaching Hospital (KATH), over a one-year period using a structured questionnaire. The gathered data were then entered into an electronic database and then analysed with SPSS version 20.0. Results: The incidence of injury following auto-tricycle crashes over the one-year period was 5.9% (95% CI: 4.9% - 7.0%) with a case fatality rate (FR) of 3.8% (95% CI: 1.3% - 8.7%). All the mortalities resulted from head and neck injuries and none of the patients involved wore a crash helmet. Only 5% of those studied wore crash helmets and were all drivers. Closed fractures accounted for 58% of the injuries, followed by open fractures, 28%. The most commonly fractured bones were the tibia/fibula, followed by the femur and then radius/ulna. The most common mechanism of injury was auto-tricycle toppling over (29%). Passengers were the most injured (48%), followed by drivers (37%) and pedestrians (15%). Most (72%) injuries among participants involved a single body part. On the injury severity scale, most (61%) of patients had minor trauma and 38% had major trauma. Conclusion: Auto-tricycle crashes account for 5.9% of injuries at the study site with a case fatality rate of 3.8%. Passengers had a higher injury rate (48%) than drivers (37%). Fractures of the tibia/fibula were most commonly associated with auto-tricycle crashes. Injuries to the head and neck were responsible for the deaths in the study participants and non-use of a crash helmet was associated with mortalities.展开更多
This paper examines the relationship between fatal road traffic accidents and potential predictors using multilayer perceptron artificial neural network (MLANN) models. The initial analysis employed twelve potential p...This paper examines the relationship between fatal road traffic accidents and potential predictors using multilayer perceptron artificial neural network (MLANN) models. The initial analysis employed twelve potential predictors, including traffic volume, prevailing weather conditions, roadway characteristics and features, drivers’ age and gender, and number of lanes. Based on the output of the model and the variables’ importance factors, seven significant variables are identified and used for further analysis to improve the performance of models. The model is optimized by systematically changing the parameters, including the number of hidden layers and the activation function of both the hidden and output layers. The performances of the MLANN models are evaluated using the percentage of the achieved accuracy, R-squared, and Sum of Square Error (SSE) functions.展开更多
Historical roadway safety analyses have used labor and time-intensive crash data collection procedures. However, crash reporting is often delayed and crash locations are reported with varying levels of spatial accurac...Historical roadway safety analyses have used labor and time-intensive crash data collection procedures. However, crash reporting is often delayed and crash locations are reported with varying levels of spatial accuracy and detail. Recent advances in connected vehicle (CV) data provide an opportunity for stakeholders to proactively identify areas of safety concerns in near-real time with high spatial precision. Public and private sector stakeholders including automotive original equipment manufacturers (OEM) and insurance providers may independently define acceleration thresholds for reporting unsafe driver behavior. Although some OEMs have provided fixed threshold hard-braking event data for a number of years, this varies by OEM and there is no published literature on the best thresholds to use for identifying emerging safety issues. This research proposes a methodology to estimate deceleration events from raw CV trajectory data at varying thresholds that can be scaled to any CV. The estimated deceleration events and crash incident records around 629 interstate exits in Indiana were analyzed for a three-month period from March 1-May 31, 2023. Nearly 20 million estimated deceleration events and 4800 crash records were spatially joined to a 2-mile search radius around each exit ramp. Results showed that deceleration events between -0.5 g and -0.4 g had the highest correlation with an R<sup>2</sup> of 0.69. This study also identifies the top 20 interstate exit locations with highest deceleration events. The framework presented in this study enables agencies and transportation professionals to perform safety evaluations on raw trajectory data without the need to integrate external data sources.展开更多
Near crash events are often regarded as an excellent surrogate measure for traffic safety research because they include abrupt changes in vehicle kinematics that can lead to deadly accident scenarios. In this paper, w...Near crash events are often regarded as an excellent surrogate measure for traffic safety research because they include abrupt changes in vehicle kinematics that can lead to deadly accident scenarios. In this paper, we introduced machine learning and deep learning algorithms for predicting near crash events using LiDAR data at a signalized intersection. To predict a near crash occurrence, we used essential vehicle kinematic variables such as lateral and longitudinal velocity, yaw, tracking status of LiDAR, etc. A deep learning hybrid model Convolutional Gated Recurrent Neural Network (CNN + GRU) was introduced, and comparative performances were evaluated with multiple machine learning classification models such as Logistic Regression, K Nearest Neighbor, Decision Tree, Random Forest, Adaptive Boost, and deep learning models like Long Short-Term Memory (LSTM). As vehicle kinematics changes occur after sudden brake, we considered average deceleration and kinematic energy drop as thresholds to identify near crashes after vehicle braking time . We looked at the next 3 seconds of this braking time as our prediction horizon. All models work best in the next 1-second prediction horizon to braking time. The results also reveal that our hybrid model gathers the greatest near crash information while working flawlessly. In comparison to existing models for near crash prediction, our hybrid Convolutional Gated Recurrent Neural Network model has 100% recall, 100% precision, and 100% F1-score: accurately capturing all near crashes. This prediction performance outperforms previous baseline models in forecasting near crash events and provides opportunities for improving traffic safety via Intelligent Transportation Systems (ITS).展开更多
This study investigates relationships between congestion and travel time performance metrics and crashes on road segments. The study focuses on work zone routes in Iowa, utilizing 2021 commercially-available probe veh...This study investigates relationships between congestion and travel time performance metrics and crashes on road segments. The study focuses on work zone routes in Iowa, utilizing 2021 commercially-available probe vehicle data and crash data. Travel time performance metrics were derived from the probe vehicle data, and crash counts were obtained from the crash data. Additional variables included road characteristics (traffic volume, road type, segment length) and a categorical variable for the presence of a work zone. A mixed effect linear regression model was employed to identify relationships between road segment crash counts and the selected performance metrics. This was accomplished for two sets of models that include congestion performance measures at different defining threshold values, along with travel time performance measures. The study results indicate that the congestion indicators, certain travel time performance measures, and traffic counts were statistically significant and positively correlated with crash counts. Indicator variables for rural interstate locations and non-active work zones have a stronger influence on crash count than those for municipal interstate locations and active work zones. These findings can inform decision-makers on work zone safety strategies and crash mitigation planning, especially in high traffic volume areas prone to congestion and queues.展开更多
深入分析交通事故数据可以为规避事故发生、降低事故严重程度提供重要理论依据,然而,在事故数据采集、传输、存储过程中往往会产生数据缺失,导致统计分析结果的准确性下降、模型的误判风险上升。本文以芝加哥2016—2021年的101452条追...深入分析交通事故数据可以为规避事故发生、降低事故严重程度提供重要理论依据,然而,在事故数据采集、传输、存储过程中往往会产生数据缺失,导致统计分析结果的准确性下降、模型的误判风险上升。本文以芝加哥2016—2021年的101452条追尾事故数据为研究对象,将原始数据按照7∶3随机分为训练集和测试集。在训练集数据上,利用生成式插补网络(Generative Adversarial Imputation Network,GAIN)实现对缺失数据的填补。为对比不同数据填补方法的效果,同时选择多重插补(Multiple Imputation by Chained Equations,MICE)算法、期望最大化(Expectation Maximization,EM)填充算法、缺失森林(MissForest)算法和K最近邻(K-Nearest Neighbor,KNN)算法对同一数据集进行数据填补,并基于填补前后变量方差变化比较不同填补算法对数据变异性的影响。在完成数据填补的基础上,构建LightGBM三分类事故严重程度影响因素分析模型。使用原始训练集数据,以及填补后的训练集数据分别训练模型,并使用未经填补的测试集数据检验模型预测效果。结果表明,经缺失值填补后,模型性能得到一定改善,使用GAIN填补数据集训练的模型,相较于原始数据训练的模型,准确率提高了6.84%,F1提高了4.61%,AUC(Area Under the Curve)提高了10.09%,且改善效果优于其他4种填补方法。展开更多
文摘Purpose: The aim of this study was to determine the incidence and pattern of injuries resulting from auto-tricycle crashes among patients in a tertiary referral centre in Ghana. Methods: Data were retrospectively extracted from hospital records of patients who got involved in auto-tricycle crashes and presented to the Accident and Emergency Centre of the Komfo Anokye Teaching Hospital (KATH), over a one-year period using a structured questionnaire. The gathered data were then entered into an electronic database and then analysed with SPSS version 20.0. Results: The incidence of injury following auto-tricycle crashes over the one-year period was 5.9% (95% CI: 4.9% - 7.0%) with a case fatality rate (FR) of 3.8% (95% CI: 1.3% - 8.7%). All the mortalities resulted from head and neck injuries and none of the patients involved wore a crash helmet. Only 5% of those studied wore crash helmets and were all drivers. Closed fractures accounted for 58% of the injuries, followed by open fractures, 28%. The most commonly fractured bones were the tibia/fibula, followed by the femur and then radius/ulna. The most common mechanism of injury was auto-tricycle toppling over (29%). Passengers were the most injured (48%), followed by drivers (37%) and pedestrians (15%). Most (72%) injuries among participants involved a single body part. On the injury severity scale, most (61%) of patients had minor trauma and 38% had major trauma. Conclusion: Auto-tricycle crashes account for 5.9% of injuries at the study site with a case fatality rate of 3.8%. Passengers had a higher injury rate (48%) than drivers (37%). Fractures of the tibia/fibula were most commonly associated with auto-tricycle crashes. Injuries to the head and neck were responsible for the deaths in the study participants and non-use of a crash helmet was associated with mortalities.
文摘This paper examines the relationship between fatal road traffic accidents and potential predictors using multilayer perceptron artificial neural network (MLANN) models. The initial analysis employed twelve potential predictors, including traffic volume, prevailing weather conditions, roadway characteristics and features, drivers’ age and gender, and number of lanes. Based on the output of the model and the variables’ importance factors, seven significant variables are identified and used for further analysis to improve the performance of models. The model is optimized by systematically changing the parameters, including the number of hidden layers and the activation function of both the hidden and output layers. The performances of the MLANN models are evaluated using the percentage of the achieved accuracy, R-squared, and Sum of Square Error (SSE) functions.
文摘Historical roadway safety analyses have used labor and time-intensive crash data collection procedures. However, crash reporting is often delayed and crash locations are reported with varying levels of spatial accuracy and detail. Recent advances in connected vehicle (CV) data provide an opportunity for stakeholders to proactively identify areas of safety concerns in near-real time with high spatial precision. Public and private sector stakeholders including automotive original equipment manufacturers (OEM) and insurance providers may independently define acceleration thresholds for reporting unsafe driver behavior. Although some OEMs have provided fixed threshold hard-braking event data for a number of years, this varies by OEM and there is no published literature on the best thresholds to use for identifying emerging safety issues. This research proposes a methodology to estimate deceleration events from raw CV trajectory data at varying thresholds that can be scaled to any CV. The estimated deceleration events and crash incident records around 629 interstate exits in Indiana were analyzed for a three-month period from March 1-May 31, 2023. Nearly 20 million estimated deceleration events and 4800 crash records were spatially joined to a 2-mile search radius around each exit ramp. Results showed that deceleration events between -0.5 g and -0.4 g had the highest correlation with an R<sup>2</sup> of 0.69. This study also identifies the top 20 interstate exit locations with highest deceleration events. The framework presented in this study enables agencies and transportation professionals to perform safety evaluations on raw trajectory data without the need to integrate external data sources.
文摘Near crash events are often regarded as an excellent surrogate measure for traffic safety research because they include abrupt changes in vehicle kinematics that can lead to deadly accident scenarios. In this paper, we introduced machine learning and deep learning algorithms for predicting near crash events using LiDAR data at a signalized intersection. To predict a near crash occurrence, we used essential vehicle kinematic variables such as lateral and longitudinal velocity, yaw, tracking status of LiDAR, etc. A deep learning hybrid model Convolutional Gated Recurrent Neural Network (CNN + GRU) was introduced, and comparative performances were evaluated with multiple machine learning classification models such as Logistic Regression, K Nearest Neighbor, Decision Tree, Random Forest, Adaptive Boost, and deep learning models like Long Short-Term Memory (LSTM). As vehicle kinematics changes occur after sudden brake, we considered average deceleration and kinematic energy drop as thresholds to identify near crashes after vehicle braking time . We looked at the next 3 seconds of this braking time as our prediction horizon. All models work best in the next 1-second prediction horizon to braking time. The results also reveal that our hybrid model gathers the greatest near crash information while working flawlessly. In comparison to existing models for near crash prediction, our hybrid Convolutional Gated Recurrent Neural Network model has 100% recall, 100% precision, and 100% F1-score: accurately capturing all near crashes. This prediction performance outperforms previous baseline models in forecasting near crash events and provides opportunities for improving traffic safety via Intelligent Transportation Systems (ITS).
文摘This study investigates relationships between congestion and travel time performance metrics and crashes on road segments. The study focuses on work zone routes in Iowa, utilizing 2021 commercially-available probe vehicle data and crash data. Travel time performance metrics were derived from the probe vehicle data, and crash counts were obtained from the crash data. Additional variables included road characteristics (traffic volume, road type, segment length) and a categorical variable for the presence of a work zone. A mixed effect linear regression model was employed to identify relationships between road segment crash counts and the selected performance metrics. This was accomplished for two sets of models that include congestion performance measures at different defining threshold values, along with travel time performance measures. The study results indicate that the congestion indicators, certain travel time performance measures, and traffic counts were statistically significant and positively correlated with crash counts. Indicator variables for rural interstate locations and non-active work zones have a stronger influence on crash count than those for municipal interstate locations and active work zones. These findings can inform decision-makers on work zone safety strategies and crash mitigation planning, especially in high traffic volume areas prone to congestion and queues.
文摘深入分析交通事故数据可以为规避事故发生、降低事故严重程度提供重要理论依据,然而,在事故数据采集、传输、存储过程中往往会产生数据缺失,导致统计分析结果的准确性下降、模型的误判风险上升。本文以芝加哥2016—2021年的101452条追尾事故数据为研究对象,将原始数据按照7∶3随机分为训练集和测试集。在训练集数据上,利用生成式插补网络(Generative Adversarial Imputation Network,GAIN)实现对缺失数据的填补。为对比不同数据填补方法的效果,同时选择多重插补(Multiple Imputation by Chained Equations,MICE)算法、期望最大化(Expectation Maximization,EM)填充算法、缺失森林(MissForest)算法和K最近邻(K-Nearest Neighbor,KNN)算法对同一数据集进行数据填补,并基于填补前后变量方差变化比较不同填补算法对数据变异性的影响。在完成数据填补的基础上,构建LightGBM三分类事故严重程度影响因素分析模型。使用原始训练集数据,以及填补后的训练集数据分别训练模型,并使用未经填补的测试集数据检验模型预测效果。结果表明,经缺失值填补后,模型性能得到一定改善,使用GAIN填补数据集训练的模型,相较于原始数据训练的模型,准确率提高了6.84%,F1提高了4.61%,AUC(Area Under the Curve)提高了10.09%,且改善效果优于其他4种填补方法。
