Tianzheng Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Tianzheng Wei has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
Tong Zhu
School of Transportation Engineering, Chang’an University, Xi’an, China
8 shared publicationsHaoxue Liu
School of Automobile, Chang’an University, Xi’an, Shaanxi, China
6 shared publicationsHan Bai
Affiliation School of Transportation and Logistics Engineering, Shandong Jiaotong University, Jinan, China
4 shared publicationsLiang Zhao
Affiliation School of Transportation and Logistics Engineering, Shandong Jiaotong University, Jinan, China
4 shared publicationsXiuguang Wang
Affiliation Shandong Zhengqu Traffic Engineering Co., Ltd, Jinan, China
3 shared publicationsChenxin Li
Affiliation Patent Examination Cooperation Sichuan Center of the Patent Office, CNIPA Sichuan Chengdu, Chengdu, China
2 shared publications
Latest Publications
- Open Access
- 23 October 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (10) , e0360312
Research on expressway traffic accident risk prediction method considering Spatio-Temporal effects
Traffic accidents arise from the dynamic evolution of traffic flow under spatio-temporal effects, and their occurrence and severity are significantly influenced by temporal and spatial variations in traffic conditions. Traditional studies predominantly rely on static accident-time data, which limits prediction accuracy. To address this gap, this study proposes a deep learning framework that integrates data preprocessing, spatio-temporal effect analysis, PCA-FCM-based risk labeling, and a CNN-BiLSTM hybrid model. Compared with CNN-LSTM, the proposed model replaces unidirectional LSTM with bidirectional LSTM to capture both forward and backward temporal dependencies. Based on the HighD naturalistic trajectory dataset, the CNN-BiLSTM model achieves an MSE of 0.362, RMSE of 0.601, MAE of 0.197 and MAPE of 0.044, outperforming LSTM, BiLSTM and CNN-LSTM with moderate running time. In accident risk classification, it reaches an overall accuracy of 90%, with class-level precision of 0.90–0.93 and F1 scores of 0.82–0.96. The results demonstrate that the proposed method effectively captures spatio-temporal dynamic characteristics and provides an efficient tool for expressway accident risk prediction.
- 23 October 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (10) , e0360312
Open AccessResearch on expressway traffic accident risk prediction method considering Spatio-Temporal effects
Traffic accidents arise from the dynamic evolution of traffic flow under spatio-temporal effects, and their occurrence and severity are significantly influenced by temporal and spatial variations in traffic conditions. Traditional studies predominantly rely on static accident-time data, which limits prediction accuracy. To address this gap, this study proposes a deep learning framework that integrates data preprocessing, spatio-temporal effect analysis, PCA-FCM-based risk labeling, and a CNN-BiLSTM hybrid model. Compared with CNN-LSTM, the proposed model replaces unidirectional LSTM with bidirectional LSTM to capture both forward and backward temporal dependencies. Based on the HighD naturalistic trajectory dataset, the CNN-BiLSTM model achieves an MSE of 0.362, RMSE of 0.601, MAE of 0.197 and MAPE of 0.044, outperforming LSTM, BiLSTM and CNN-LSTM with moderate running time. In accident risk classification, it reaches an overall accuracy of 90%, with class-level precision of 0.90–0.93 and F1 scores of 0.82–0.96. The results demonstrate that the proposed method effectively captures spatio-temporal dynamic characteristics and provides an efficient tool for expressway accident risk prediction.
- 7 April 2026
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
A study on driver behavior differences in emergency rear-end collision scenarios
Driver behavior in car-following scenarios is highly scenario-dependent. To explore the effect of different vehicle, rear-end collision urgent situations on driver behaviors, this study designed the experimental scenarios for the urgency of vehicle rear-end collisions based on virtual simulation technology. Based on the kinematics parameters of the two vehicles and the driver’s behavior, the driving behavior under different rear-end collision conflict risks was analyzed, and a rear-end collision risk prediction model was constructed by using the binary Logistic regression method. Meanwhile, the influence of the two-factor interaction effect on the risk of rear-end collision was explored by using XGBoost and SHAP methods. The results show that with the increase of collision urgency, the brake reaction time (BRT) is shortened, and the maximum deceleration and deceleration rate increase significantly ( p < 0.05). Under the same conflict urgency scenario, the higher speed of the lead vehicle will make a shorter BRT and following spacing, the maximum deceleration and the deceleration rate significantly increase ( p < 0.05). With the increase of braking deceleration in the lead vehicle, the BRT decreases significantly ( p < 0.05), but the following spacing and the maximum deceleration increase significantly ( p < 0.05). With the decrease in time to collision (TTC) and the increase in vehicle speed, the proportion of drivers taking braking steering behavior increases, and they generally take braking measures first. The rear-end collision prediction model can accurately predict the collision risk, and the accuracy is 85.2%. The results of this study could provide a reference for the further improvement of road traffic safety management and in-vehicle assistance systems.
- 7 April 2026
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
A study on driver behavior differences in emergency rear-end collision scenarios
Driver behavior in car-following scenarios is highly scenario-dependent. To explore the effect of different vehicle, rear-end collision urgent situations on driver behaviors, this study designed the experimental scenarios for the urgency of vehicle rear-end collisions based on virtual simulation technology. Based on the kinematics parameters of the two vehicles and the driver’s behavior, the driving behavior under different rear-end collision conflict risks was analyzed, and a rear-end collision risk prediction model was constructed by using the binary Logistic regression method. Meanwhile, the influence of the two-factor interaction effect on the risk of rear-end collision was explored by using XGBoost and SHAP methods. The results show that with the increase of collision urgency, the brake reaction time (BRT) is shortened, and the maximum deceleration and deceleration rate increase significantly ( p < 0.05). Under the same conflict urgency scenario, the higher speed of the lead vehicle will make a shorter BRT and following spacing, the maximum deceleration and the deceleration rate significantly increase ( p < 0.05). With the increase of braking deceleration in the lead vehicle, the BRT decreases significantly ( p < 0.05), but the following spacing and the maximum deceleration increase significantly ( p < 0.05). With the decrease in time to collision (TTC) and the increase in vehicle speed, the proportion of drivers taking braking steering behavior increases, and they generally take braking measures first. The rear-end collision prediction model can accurately predict the collision risk, and the accuracy is 85.2%. The results of this study could provide a reference for the further improvement of road traffic safety management and in-vehicle assistance systems.
- 28 April 2025
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 240 (4) , 1887-1899
Effects of IVIS touchscreen operation tasks on driver’s mental workload based on explainable CatBoost algorithm
With the widespread adoption of touchscreen-based in-vehicle information systems (IVIS) in vehicles, a large amount of effective information is displayed on the system, resulting in a significant increase in the frequency of driver interaction with these systems. However, complex operational tasks can lead to elevated mental workload, thereby impacting driving safety. This study aims to investigate the effects of secondary tasks involving IVIS touchscreen operations on driver mental workload. Through driving simulation experiments and survey questionnaires, three car-following scenarios were designed at speed levels of 60, 40, and 20 km/h. A total of 36 participants completed the IVIS secondary task driving simulation tests. Using statistical analysis and interpretable machine learning methods, a driver mental workload prediction model based on the CatBoost algorithm was constructed. Shapley Additive exPlanations (SHAP), partial dependence plot (PDP), and individual conditional expectation (ICE) were used to comprehensively analyze the relationship between important driving behavior characteristics and mental workload. The results of the study indicate that as the number of manual operations of IVIS touchscreen secondary tasks increases, the driver’s mental workload, standard deviation of speed, standard deviation of Lateral offset distance, task completion time, and saccade number correspondingly significant increase ( p < 0.05). With the increase of the driver’s mental workload, the speed of the following vehicle decreases ( p < 0.01), along with a significant reduction in following distance ( p < 0.05). When the number of IVIS secondary task manual operations exceeded 3, the probability of a high mental workload significantly increased. These findings provide a basis for designing safer IVIS, contributing to enhanced driving safety and improved driver experience, holding significant theoretical and practical application value.
- 28 April 2025
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 240 (4) , 1887-1899
Effects of IVIS touchscreen operation tasks on driver’s mental workload based on explainable CatBoost algorithm
With the widespread adoption of touchscreen-based in-vehicle information systems (IVIS) in vehicles, a large amount of effective information is displayed on the system, resulting in a significant increase in the frequency of driver interaction with these systems. However, complex operational tasks can lead to elevated mental workload, thereby impacting driving safety. This study aims to investigate the effects of secondary tasks involving IVIS touchscreen operations on driver mental workload. Through driving simulation experiments and survey questionnaires, three car-following scenarios were designed at speed levels of 60, 40, and 20 km/h. A total of 36 participants completed the IVIS secondary task driving simulation tests. Using statistical analysis and interpretable machine learning methods, a driver mental workload prediction model based on the CatBoost algorithm was constructed. Shapley Additive exPlanations (SHAP), partial dependence plot (PDP), and individual conditional expectation (ICE) were used to comprehensively analyze the relationship between important driving behavior characteristics and mental workload. The results of the study indicate that as the number of manual operations of IVIS touchscreen secondary tasks increases, the driver’s mental workload, standard deviation of speed, standard deviation of Lateral offset distance, task completion time, and saccade number correspondingly significant increase ( p < 0.05). With the increase of the driver’s mental workload, the speed of the following vehicle decreases ( p < 0.01), along with a significant reduction in following distance ( p < 0.05). When the number of IVIS secondary task manual operations exceeded 3, the probability of a high mental workload significantly increased. These findings provide a basis for designing safer IVIS, contributing to enhanced driving safety and improved driver experience, holding significant theoretical and practical application value.
- 24 April 2025
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2679 (6) , 749-769
Study on Mixed Traffic Flow Characteristics in an Expressway Environment Considering Illuminance Effects
This study examines the influence of illuminance on car-following characteristics in mixed traffic flow. We selected an expressway tunnel with a large variation in illuminance and designed a real vehicle test to explore the influence of changes in tunnel illuminance on car-following characteristics. The load factor of vehicle recognition based on the critical speed of illuminance change is introduced to correct the delay time of adaptive cruise control vehicle recognition. The continuous cellular automata model was used to simulate a mixed traffic flow. The characteristics of the mixed traffic flow under the influence of illuminance are analyzed for traffic efficiency, stability, and safety. The results show that changes in illuminance cause vehicle speeds to diminish at tunnel entrances and exits while following distances increase. Inside the tunnel, however, vehicle speeds and following distances remain relatively stable. As connected autonomous vehicles (CAV) become more common, the maximum flow and average speed in expressway tunnels increase. In addition, the time to collision (TTC) value increases gradually, while the time exposed rear collision risk index (TERCRI) decreases gradually. When CAV penetration exceeds 80%, the speed-density curve is smoother, the vehicle motion state is close to a straight line on a certain slope, and the position of deceleration is less, which is close to the ideal state of stable driving. This study could have a positive impact in reducing the speed fluctuation of vehicles in expressway tunnels, improving the efficiency of traffic, and preventing rear-end crashes in mixed traffic flow environments.
- 24 April 2025
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2679 (6) , 749-769
Study on Mixed Traffic Flow Characteristics in an Expressway Environment Considering Illuminance Effects
This study examines the influence of illuminance on car-following characteristics in mixed traffic flow. We selected an expressway tunnel with a large variation in illuminance and designed a real vehicle test to explore the influence of changes in tunnel illuminance on car-following characteristics. The load factor of vehicle recognition based on the critical speed of illuminance change is introduced to correct the delay time of adaptive cruise control vehicle recognition. The continuous cellular automata model was used to simulate a mixed traffic flow. The characteristics of the mixed traffic flow under the influence of illuminance are analyzed for traffic efficiency, stability, and safety. The results show that changes in illuminance cause vehicle speeds to diminish at tunnel entrances and exits while following distances increase. Inside the tunnel, however, vehicle speeds and following distances remain relatively stable. As connected autonomous vehicles (CAV) become more common, the maximum flow and average speed in expressway tunnels increase. In addition, the time to collision (TTC) value increases gradually, while the time exposed rear collision risk index (TERCRI) decreases gradually. When CAV penetration exceeds 80%, the speed-density curve is smoother, the vehicle motion state is close to a straight line on a certain slope, and the position of deceleration is less, which is close to the ideal state of stable driving. This study could have a positive impact in reducing the speed fluctuation of vehicles in expressway tunnels, improving the efficiency of traffic, and preventing rear-end crashes in mixed traffic flow environments.
- 27 February 2025
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 240 (2-3) , 1165-1178
Impact of driving style on rear-end crash risk in leading vehicle emergency braking situations
The driving style and skill of drivers have a significant impact on traffic safety, and the crash urgency level of the situation can affect drivers’ behavior. To investigate the differences in evasion behavior among drivers with different driving styles in rear-end crash urgency scenarios, this study classified drivers into aggressive and conservative types based on risk tendencies obtained from the Iowa Gambling Task. Using simulated driving equipment, rear-end crash risk scenarios were classified based on time to collision (TTC) thresholds and drivers’ reaction behavior characteristics in emergency braking situations were collected. Comparative analysis of evasion behavior characteristics among drivers with different styles was conducted, and the influence trends of various factors on the risk of vehicle crash events were explored using binary logistic regression. The results showed that with increasing crash urgency, drivers’ braking reaction time (BRT) decreased, while vehicle speed and jerk increased. Novice drivers exhibited longer BRT, larger vehicle spacing, higher maximum braking deceleration, jerk, and TTC −1 compared to experienced drivers. Aggressive drivers had significantly longer BRT than conservative drivers, while maximum braking deceleration and jerk were significantly lower. Vehicle spacing, maximum braking deceleration, and jerk were negatively correlated with crash risk. This study provides important references for the analysis of evasion behavior among different driving styles in crash urgency scenarios and the design of active safety intervention methods.
- 27 February 2025
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 240 (2-3) , 1165-1178
Impact of driving style on rear-end crash risk in leading vehicle emergency braking situations
The driving style and skill of drivers have a significant impact on traffic safety, and the crash urgency level of the situation can affect drivers’ behavior. To investigate the differences in evasion behavior among drivers with different driving styles in rear-end crash urgency scenarios, this study classified drivers into aggressive and conservative types based on risk tendencies obtained from the Iowa Gambling Task. Using simulated driving equipment, rear-end crash risk scenarios were classified based on time to collision (TTC) thresholds and drivers’ reaction behavior characteristics in emergency braking situations were collected. Comparative analysis of evasion behavior characteristics among drivers with different styles was conducted, and the influence trends of various factors on the risk of vehicle crash events were explored using binary logistic regression. The results showed that with increasing crash urgency, drivers’ braking reaction time (BRT) decreased, while vehicle speed and jerk increased. Novice drivers exhibited longer BRT, larger vehicle spacing, higher maximum braking deceleration, jerk, and TTC −1 compared to experienced drivers. Aggressive drivers had significantly longer BRT than conservative drivers, while maximum braking deceleration and jerk were significantly lower. Vehicle spacing, maximum braking deceleration, and jerk were negatively correlated with crash risk. This study provides important references for the analysis of evasion behavior among different driving styles in crash urgency scenarios and the design of active safety intervention methods.
- 28 January 2025
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 240 (1) , 573-589
Impact of touchscreen-based in-vehicle information systems operational complexity on driving behavior and mental workload in car following conditions
The interaction between the driver and the in-vehicle information system (IVIS) significantly impacts both the driver’s behavior and driving safety. This research aims to explore the influence of task complexity in touchscreen-based IVIS on driving behavior and mental workload. Based on the driving simulation, a test was designed for drivers to manually operate the IVIS under car-following conditions. Driving behavior and eye movement data were collected under different manual touchscreen task complexities. The effects of driver gender, driving experience, and the number of hands-on operations on driving behavior and mental workload were investigated by analysis of variance. A quantitative model of driver mental workload was constructed using exploratory factor analysis. The validity of this model was confirmed through the NASA-TLX scale. Additionally, a driver mental workload discrimination model was constructed using the Bayes discrimination method to achieve an accurate classification of mental workload. The results show that IVIS touchscreen operation tasks significantly affect driving behavior. The quantitative model, based on driving behavior and eye movement indicators, facilitates the evaluation of the driver’s mental workload under the influence of manual IVIS touchscreen operation tasks. Variables such as the driver’s gender, driving experience, and the number of hands-on operations significantly affect mental workload. The driver mental workload discrimination model correctly discriminated 87.0% of driver test samples. Cross-validation results showed an accuracy rate of 84.30%, indicating the model’s high precision in discriminating mental workload.
- 28 January 2025
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 240 (1) , 573-589
Impact of touchscreen-based in-vehicle information systems operational complexity on driving behavior and mental workload in car following conditions
The interaction between the driver and the in-vehicle information system (IVIS) significantly impacts both the driver’s behavior and driving safety. This research aims to explore the influence of task complexity in touchscreen-based IVIS on driving behavior and mental workload. Based on the driving simulation, a test was designed for drivers to manually operate the IVIS under car-following conditions. Driving behavior and eye movement data were collected under different manual touchscreen task complexities. The effects of driver gender, driving experience, and the number of hands-on operations on driving behavior and mental workload were investigated by analysis of variance. A quantitative model of driver mental workload was constructed using exploratory factor analysis. The validity of this model was confirmed through the NASA-TLX scale. Additionally, a driver mental workload discrimination model was constructed using the Bayes discrimination method to achieve an accurate classification of mental workload. The results show that IVIS touchscreen operation tasks significantly affect driving behavior. The quantitative model, based on driving behavior and eye movement indicators, facilitates the evaluation of the driver’s mental workload under the influence of manual IVIS touchscreen operation tasks. Variables such as the driver’s gender, driving experience, and the number of hands-on operations significantly affect mental workload. The driver mental workload discrimination model correctly discriminated 87.0% of driver test samples. Cross-validation results showed an accuracy rate of 84.30%, indicating the model’s high precision in discriminating mental workload.
- 27 July 2024
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2679 (1) , 2166-2182
Effects of Driver Gender, Driving Experience, and Visibility on Car-Following Behavior
Weather visibility interference has a significant impact on driver car-following behavior. To investigate drivers’ car-following behavior and emergency avoidance behaviors under different visibility disturbances, scenarios are constructed under different foggy concentration environments based on driving simulation, and the drivers’ response behaviors are collected in the stable car-following state and emergency rear-end scenarios. Exploring the differential effects of gender and driving experience on driving behavior for fog concentrations based on multifactorial analysis of variance. A quantitative model of car-following risk is constructed based on factor analysis, and a linear mixed model is used to explore the comprehensive effects of fog concentration, speed, and the following distance at the braking time on drivers’ braking reaction time by fully considering the differences in individual behaviors. The results show that driving behavior is significantly affected by visual visibility, driver’s gender, and driving experience. With the decrease of visibility, following driving speed decreases, the following distance is shortened, the headway decreases, and the standard deviation of lane lateral offset distance increases. The rear-end collision risk of an experienced driver is higher than that of a novice driver, and the rear-end collision risk of the female is higher than that of male. The risk of collision is higher when traveling in light fog. In emergency rear-end collision scenarios, as visibility decreases, braking reaction time increases, and the risk of collision conflict increases at the moment of driver braking. The braking reaction time of the driver decreases with the increase of the speed and increases with the increase of the distance when the front vehicle is braking. The results of this study provide theoretical support and technical reference for effectively improving driving safety in a bad-visibility environment.
- 27 July 2024
- Published by SAGE Publications in Transportation Research Record: Journal of the Transportation Research Board
- Vol. 2679 (1) , 2166-2182
Effects of Driver Gender, Driving Experience, and Visibility on Car-Following Behavior
Weather visibility interference has a significant impact on driver car-following behavior. To investigate drivers’ car-following behavior and emergency avoidance behaviors under different visibility disturbances, scenarios are constructed under different foggy concentration environments based on driving simulation, and the drivers’ response behaviors are collected in the stable car-following state and emergency rear-end scenarios. Exploring the differential effects of gender and driving experience on driving behavior for fog concentrations based on multifactorial analysis of variance. A quantitative model of car-following risk is constructed based on factor analysis, and a linear mixed model is used to explore the comprehensive effects of fog concentration, speed, and the following distance at the braking time on drivers’ braking reaction time by fully considering the differences in individual behaviors. The results show that driving behavior is significantly affected by visual visibility, driver’s gender, and driving experience. With the decrease of visibility, following driving speed decreases, the following distance is shortened, the headway decreases, and the standard deviation of lane lateral offset distance increases. The rear-end collision risk of an experienced driver is higher than that of a novice driver, and the rear-end collision risk of the female is higher than that of male. The risk of collision is higher when traveling in light fog. In emergency rear-end collision scenarios, as visibility decreases, braking reaction time increases, and the risk of collision conflict increases at the moment of driver braking. The braking reaction time of the driver decreases with the increase of the speed and increases with the increase of the distance when the front vehicle is braking. The results of this study provide theoretical support and technical reference for effectively improving driving safety in a bad-visibility environment.
- 3 March 2024
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 239 (6) , 1902-1916
Analysis of following vehicle driver injury severity in rear-end collision on straight road based on LightGBM and SHAP
Rear-end accidents, as one of the common accident types, may cause serious injuries to the driver and passengers. There are mutual coupling effects among causal features of rear-end accidents, and traditional analytical methods may lead to modeling distortion due to assumption constraints. In this study, the Light Gradient Boosting Machine (LightGBM), a machine learning algorithm, is used to model the injury severity of the following vehicle driver in a rear-end crash accident on a straight road, based on the data from the China In-depth Accident Study data. The Shapley Additive Explanations (SHAP) method was used to interpret the results of the LightGBM model and to analyze the relationship between factors and driver injury severity. The results show that location familiarity, willingness to take risks, and driving time have a significant impact on the injury severity for following vehicle drivers involved in rear-end crashes. Cloudy conditions increase a driver’s risk of being involved in a fatal or injured rear-end collision. In rain, snow, hail, and foggy conditions, have a higher propensity to cause driver fatalities in the crash event. The rear-end crash of passenger cars resulted in a higher death probability of the following vehicle driver compared to sedan and truck, and female drivers are more likely to be involved in uninjured accidents compared to males. These results are informative for preventing rear-end accidents and reducing the extent of accidental injuries.
- 3 March 2024
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 239 (6) , 1902-1916
Analysis of following vehicle driver injury severity in rear-end collision on straight road based on LightGBM and SHAP
Rear-end accidents, as one of the common accident types, may cause serious injuries to the driver and passengers. There are mutual coupling effects among causal features of rear-end accidents, and traditional analytical methods may lead to modeling distortion due to assumption constraints. In this study, the Light Gradient Boosting Machine (LightGBM), a machine learning algorithm, is used to model the injury severity of the following vehicle driver in a rear-end crash accident on a straight road, based on the data from the China In-depth Accident Study data. The Shapley Additive Explanations (SHAP) method was used to interpret the results of the LightGBM model and to analyze the relationship between factors and driver injury severity. The results show that location familiarity, willingness to take risks, and driving time have a significant impact on the injury severity for following vehicle drivers involved in rear-end crashes. Cloudy conditions increase a driver’s risk of being involved in a fatal or injured rear-end collision. In rain, snow, hail, and foggy conditions, have a higher propensity to cause driver fatalities in the crash event. The rear-end crash of passenger cars resulted in a higher death probability of the following vehicle driver compared to sedan and truck, and female drivers are more likely to be involved in uninjured accidents compared to males. These results are informative for preventing rear-end accidents and reducing the extent of accidental injuries.
- 29 November 2023
- Published by Taylor & Francis in Traffic Injury Prevention
- Vol. 25 (2) , 194-201
Predicting and factor analysis of rider injury severity in two-wheeled motorcycle and vehicle crash accidents based on an interpretable machine learning framework
As one of the vulnerable road users in accidents, how to improve the two-wheeled motorcyclist's driving safety and reduce accident injury is a public health issue. Accurate identification of the factors influencing the severity of accidents is an important prerequisite for mitigating injury from crashes. Based on a vehicle and a two-wheeled motorcycle crash accident data from the China in-depth accident study database (CIDAS), this study uses the performance evaluation indicators of accuracy, precision, recall, F1-score, AUC, and the ROC curve. The classification and prediction performances of the six machine learning methods on the dataset are compared, and the LightGBM algorithm with the best performance is selected to model the accident injury severity of the motorcyclists. The SHAP method is used to extend the interpretability of the LightGBM model results. Based on the SHAP method, the importance, main effect, and the interaction effect of factors under each accident injury severity are quantitatively analyzed. The model prediction accuracy is 92.6%, the F1-Score is 92.8%, and the AUC value is 0.986. The importance of factors varies with the accident injury severity of motorcyclists. The kilometers traveled per year by the driver, the throwing distance of the motorcyclist, and the road speed limit are the three most important factors. The motorcyclist is more likely to suffer fatal injuries when the throwing distance is >1,000 cm. The prediction model of driver injury severity based on LightGBM algorithm has a good prediction performance. It can be used to analyze the influence factors of injury severity in two-wheeled motorcyclist accident by combining the model with SHAP method. These results could help the traffic management department to take measures to reduce accident injury of motorcyclists.
- 29 November 2023
- Published by Taylor & Francis in Traffic Injury Prevention
- Vol. 25 (2) , 194-201
Predicting and factor analysis of rider injury severity in two-wheeled motorcycle and vehicle crash accidents based on an interpretable machine learning framework
As one of the vulnerable road users in accidents, how to improve the two-wheeled motorcyclist's driving safety and reduce accident injury is a public health issue. Accurate identification of the factors influencing the severity of accidents is an important prerequisite for mitigating injury from crashes. Based on a vehicle and a two-wheeled motorcycle crash accident data from the China in-depth accident study database (CIDAS), this study uses the performance evaluation indicators of accuracy, precision, recall, F1-score, AUC, and the ROC curve. The classification and prediction performances of the six machine learning methods on the dataset are compared, and the LightGBM algorithm with the best performance is selected to model the accident injury severity of the motorcyclists. The SHAP method is used to extend the interpretability of the LightGBM model results. Based on the SHAP method, the importance, main effect, and the interaction effect of factors under each accident injury severity are quantitatively analyzed. The model prediction accuracy is 92.6%, the F1-Score is 92.8%, and the AUC value is 0.986. The importance of factors varies with the accident injury severity of motorcyclists. The kilometers traveled per year by the driver, the throwing distance of the motorcyclist, and the road speed limit are the three most important factors. The motorcyclist is more likely to suffer fatal injuries when the throwing distance is >1,000 cm. The prediction model of driver injury severity based on LightGBM algorithm has a good prediction performance. It can be used to analyze the influence factors of injury severity in two-wheeled motorcyclist accident by combining the model with SHAP method. These results could help the traffic management department to take measures to reduce accident injury of motorcyclists.
- 7 September 2022
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 238 (1) , 172-186
Risk prediction and factor analysis of rider’s injury severity in passenger car and E-bike accidents based on interpretable machine learning
Casualties and property losses caused by the passenger car and electric bicycle crash accidents increased year by year. Assessment of the relevant risk factors of injury severity in passenger car and electric bicycle crashes could help to mitigate crash severity. This study uses an emerging machine learning method to predict the relationship between the risk factors and bicyclist accident injury severity in passenger car-electric bicycle collision accidents. The model’s performance is compared and evaluated based on accuracy, precision, recall, F1-Score, area under curve (AUC), and receiver operating characteristic curve (ROC). An interpretable machine learning framework Shapley additive explanations (SHAP) is used to further analyze the relationship between risk factors and bicyclist injury severity. It is found that we can adopt the light gradient boosting machine (LightGBM) algorithm after hyper-parameter optimization to get the highest accuracy (94.85%), precision (95.2%), recall (94.9%), F1-Score (95%), and AUC (0.993) based on the accident data of electric bicycles and passenger cars in the China in-depth accident study dataset from 2014 to 2018. The model can be used to assess new accident cases based on the model learning rate. There are some new findings in the aspects of bicyclists’ physical factors and electric vehicle characteristics. The throwing distance of the bicyclist has a positive impact on the injury severity. The bicyclist is more likely to suffer more serious injuries in a crash accident when the bicyclist is male, or shorter. Electric bicycles have a smaller handlebar width. In general, the lower handlebar height is, or lower saddle height is, the more serious the bicyclist’s injury is. Safety training for drivers can help reduce the injury severity in crash accidents and improve traffic safety.
- 7 September 2022
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 238 (1) , 172-186
Risk prediction and factor analysis of rider’s injury severity in passenger car and E-bike accidents based on interpretable machine learning
Casualties and property losses caused by the passenger car and electric bicycle crash accidents increased year by year. Assessment of the relevant risk factors of injury severity in passenger car and electric bicycle crashes could help to mitigate crash severity. This study uses an emerging machine learning method to predict the relationship between the risk factors and bicyclist accident injury severity in passenger car-electric bicycle collision accidents. The model’s performance is compared and evaluated based on accuracy, precision, recall, F1-Score, area under curve (AUC), and receiver operating characteristic curve (ROC). An interpretable machine learning framework Shapley additive explanations (SHAP) is used to further analyze the relationship between risk factors and bicyclist injury severity. It is found that we can adopt the light gradient boosting machine (LightGBM) algorithm after hyper-parameter optimization to get the highest accuracy (94.85%), precision (95.2%), recall (94.9%), F1-Score (95%), and AUC (0.993) based on the accident data of electric bicycles and passenger cars in the China in-depth accident study dataset from 2014 to 2018. The model can be used to assess new accident cases based on the model learning rate. There are some new findings in the aspects of bicyclists’ physical factors and electric vehicle characteristics. The throwing distance of the bicyclist has a positive impact on the injury severity. The bicyclist is more likely to suffer more serious injuries in a crash accident when the bicyclist is male, or shorter. Electric bicycles have a smaller handlebar width. In general, the lower handlebar height is, or lower saddle height is, the more serious the bicyclist’s injury is. Safety training for drivers can help reduce the injury severity in crash accidents and improve traffic safety.