Predicting Highway Risk Event with Trajectory Data: A Joint Approach of Traffic Flow and Vehicle Kinematics
Abstract
:1. Introduction
- The proposal of a risk-event-based method for extracting traffic flow features and inter-vehicle kinematic features designed to analyze collision risk on highways.
- The development of machine-learning-based risk identification and prediction models, specifically the Risk Identification Model, Risk Prediction Model-5s, and Risk Prediction Model-10s. These models were used to compare the performance of five distinct machine-learning approaches under various data-processing strategies.
- An exploration of the impacts of traffic flow features and inter-vehicle kinematic features on risk events, confirming the effectiveness of joint prediction using these two features.
2. Background
2.1. Identification of Risk Events
2.2. Real-Time Traffic Risk Identification and Prediction Methods
2.2.1. Data Source
2.2.2. Feature Selection
2.2.3. Classification Model
3. Data Preparation
3.1. Trajectory Dataset
3.2. Identification of Risk and Non-Risk Events
3.2.1. MTTC-Based Risk Event Identification
3.2.2. Risk Event and Non-Risk Event Extraction
3.3. Traffic Flow Features and Inter-Vehicle Kinematic Feature Extraction
3.3.1. Temporal Range
3.3.2. Feature Variable Extraction
4. Methodology
5. Results and Discussion
5.1. Variable Importance
5.2. Risk Identification Model
5.3. Risk Prediction Model
- Risk Prediction Model-5s: This model utilizes the currently extracted traffic flow features and inter-vehicle kinematic features to predict the risk situation 5 s later.
- Risk Prediction Model-10s: This model employs the currently extracted traffic flow features and inter-vehicle kinematic features to predict the risk situation 10 s later.
6. Conclusions
- The XGBoost model trained on the RENN dataset emerges as the superior model for risk identification, with an F1 score of 0.604, and can identify 53.9% of risk events with a 66.9% correct risk identification rate. However, it is important to note that the resampling strategy is not always effective when developing risk analysis models and a decision on whether to adopt a resampling strategy and to select an appropriate resampling technique needs to be made based on the characteristics of the model and the target metrics.
- The RF model demonstrated optimal performance under both risk prediction conditions, with precision and recall of 0.749 and 0.258 for the 5-s-advance scenario and 0.720 and 0.257 for the 10-s-advance scenario, respectively. In addition, the XGBoost model also achieved a strong risk prediction capability with F1 values of 0.356 and 0.361, indicating that the integrated learning model has strong fitting and generalization performance in the identification and prediction of risk.
- In the sensitivity analysis of traffic features, the model using complete features achieved higher F1 scores and AUC values compared to the model using traffic flow features or inter-vehicle kinematics features alone, indicating that the combined use of traffic flow features and inter-vehicle kinematics features yields the best.
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Abbreviations
SSM | Surrogate safety measure |
TTC | Time to collision |
MTTC | Modified Time to Collision |
DRAC | Deceleration to avoid collision |
PET | Post-encroachment time |
HIGHD | The Highway Drone Dataset |
LR | Logistic Regression |
KNN | K-Nearest Neighbors |
XGBoost | eXtreme Gradient Boosting |
RF | Random Forests |
MLP | Multilayer Perceptron |
AUC | Area Under the Receiver Operating Characteristic Curve |
SMOTE | Synthetic Minority Oversampling Technique |
RENN | Repeated Edited Nearest Neighbors |
Appendix A
- Refer to Section 3.2.2 to identify whether each trajectory contains risk events or non-risk events in turn, and obtain the frame time of event occurrence. For example, the trajectory with id 76 contains the risk event, which occurs at frame time 1509.
- Refer to Section 3.3.1 to calculate the time range for extracting traffic flow features and inter-vehicle kinematic features. Taking the above risk event as an example, the time range of feature extraction for traffic flow for risk identification is 760 to 1509 frames, and the time range of feature extraction for inter-vehicle kinematic features is 1485 to 1509 frames.
- Refer to Section 3.3.2 to calculate the traffic flow features and inter-vehicle kinematic features of the corresponding samples of events. For the traffic flow features, first, trajectory data within the time range of traffic flow feature extraction is found, and parts in the same direction are screened out. Then, the traffic flow features are calculated using the frame data that belong to the first entry or exit of the road during this period. For the inter-vehicle kinematic features, first, trajectory data within the time range of inter-vehicle kinematic feature extraction is found, and the parts in the same direction and lane are screened out. Then, the inter-vehicle kinematic features are calculated. Table A2 shows the feature extraction results of risk events in the complete dataset.
Frame | Id | Width | x | LaneId | PrecedingId | |||||
---|---|---|---|---|---|---|---|---|---|---|
1507 | 76 | 7.48 | 334.66 | 2 | −27.82 | 0.82 | −0.38 | 0.46 | 74 | 2.55 |
1508 | 76 | 7.48 | 333.54 | 2 | −27.84 | 0.84 | −0.37 | 0.45 | 74 | 2.51 |
1509 | 76 | 7.48 | 332.42 | 2 | −27.85 | 0.87 | −0.36 | 0.45 | 74 | 2.47 |
1507 | 74 | 8.49 | 314.94 | 2 | −23.65 | −0.11 | 0.2 | −0.03 | 72 | 21.15 |
1508 | 74 | 8.49 | 313.98 | 2 | −23.64 | −0.11 | 0.2 | −0.02 | 72 | 19.88 |
1509 | 74 | 8.49 | 313.01 | 2 | −23.63 | −0.11 | 0.2 | −0.01 | 72 | 19.86 |
1507 | 77 | 4.45 | 309.41 | 3 | −40.88 | 0.02 | 1.15 | 0.08 | 69 | 7.03 |
1508 | 77 | 4.45 | 307.79 | 3 | −40.84 | 0.02 | 1.15 | 0.08 | 69 | 7.02 |
1509 | 77 | 4.45 | 306.16 | 3 | −40.79 | 0.03 | 1.14 | 0.08 | 69 | 7.02 |
AvgV_U | AvgV_D | DiffV_UD | StdV_U | StdV_D | CvV_U | CvV_D | Vo_U | Vo_D |
---|---|---|---|---|---|---|---|---|
35.20 | 35.35 | 0.14 | 1.71 | 3.96 | 0.04 | 0.11 | 12.00 | 8.00 |
DiffVo_DU | Diff_AvgV_U | Diff_AvgV_D | Diff_StdV_U | Diff_StdV_D | Diff_CvV_U | Diff_CvV_D | Diff_Vo_U | Diff_Vo_D |
4.00 | 8.56 | 6.35 | 2.25 | 0.98 | 0.10 | 0.05 | 5.00 | 1.00 |
Max_XV | Max_Diff_XV | Max_YV | Max_XA | Max_Diff_XA | Max_YA | Min_D | ||
37.61 | 4.55 | 0.30 | 0.86 | 0.17 | 0.08 | 29.10 |
References
- WHO. Road Traffic Injuries. 2022. Available online: https://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries (accessed on 20 June 2022).
- WHO. Global Plan for the Decade of Action for Road Safety 2021–2030. 2021. Available online: https://www.who.int/publications/m/item/global-plan-for-the-decade-of-action-for-road-safety-2021-2030 (accessed on 20 October 2021).
- Flannagan, C.; LeBlanc, D.; Bogard, S.; Nobukawa, K.; Narayanaswamy, P.; Leslie, A.; Kiefer, R.; Marchione, M.; Beck, C.S.; Lobes, K. Large-scale field test of forward collision alert and lane departure warning systems. Natl. Acad. Sci. 2016, 01605729. [Google Scholar]
- FHWA. Highway Safety Improvement Program (HSIP). 2020. Available online: https://safety.fhwa.dot.gov/hsip/hsip.cfm (accessed on 11 February 2023).
- Yasmin, S.; Eluru, N.; Wang, L.; Abdel-Aty, M. A joint framework for static and real-time crash risk analysis. Anal. Methods Accid. Res. 2018, 18, 45–56. [Google Scholar] [CrossRef]
- Yuan, J.; Abdel-Aty, M. Approach-Level Real-Time Crash Risk Analysis for Signalized Intersections. Accid. Anal. Prev. 2018, 119, 274–289. [Google Scholar] [CrossRef]
- Lu, Q.L.; Yang, K.; Antoniou, C. Crash risk analysis for the mixed traffic flow with human-driven and connected and autonomous vehicles. In Proceedings of the 2021 IEEE International Intelligent Transportation Systems Conference (ITSC), Indianapolis, IN, USA, 19–22 September 2021; pp. 1233–1238. [Google Scholar]
- Dan Chia, W.M.; Loong Keoh, S.; Michala, A.L.; Goh, C. Real-time Recursive Risk Assessment Framework for Autonomous Vehicle Operations. In Proceedings of the 2021 IEEE 93rd Vehicular Technology Conference (VTC2021-Spring), Helsinki, Finland, 25–28 April 2021; pp. 1–7. [Google Scholar]
- Xu, C.; Tarko, A.P.; Wang, W.; Liu, P. Predicting crash likelihood and severity on freeways with real-time loop detector data. Accid. Anal. Prev. 2013, 57, 30–39. [Google Scholar] [CrossRef]
- Chen, Z.; Qin, X. A novel method for imminent crash prediction and prevention. Accid. Anal. Prev. 2019, 125, 320–329. [Google Scholar] [CrossRef]
- Liu, M.; Chen, Y. Predicting real-time crash risk for urban expressways in China. Math. Probl. Eng. 2017, 2017, 6263726. [Google Scholar] [CrossRef]
- Krajewski, R.; Bock, J.; Kloeker, L.; Eckstein, L. The highD Dataset: A Drone Dataset of Naturalistic Vehicle Trajectories on German Highways for Validation of Highly Automated Driving Systems. In Proceedings of the 2018 21st International Conference on Intelligent Transportation Systems (ITSC), Maui, HI, USA, 4–7 November 2018. [Google Scholar]
- Wang, J.; Fu, T.; Xue, J.; Li, C.; Song, H.; Xu, W.; Shangguan, Q. Realtime wide-area vehicle trajectory tracking using millimeter-wave radar sensors and the open TJRD TS dataset. Int. J. Transp. Sci. Technol. 2023, 12, 273–290. [Google Scholar] [CrossRef]
- Katrakazas, C.; Quddus, M.; Chen, W.H. A simulation study of predicting real-time conflict-prone traffic conditions. IEEE Trans. Intell. Transp. Syst. 2018, 19, 3196–3207. [Google Scholar] [CrossRef]
- Yang, D.; Wu, Y.; Sun, F.; Chen, J.; Zhai, D.; Fu, C. Freeway accident detection and classification based on the multi-vehicle trajectory data and deep learning model. Transp. Res. Part C Emerg. Technol. 2021, 130, 103303. [Google Scholar] [CrossRef]
- Guo, F.; Klauer, S.G.; McGill, M.T.; Dingus, T.A. Evaluating the Relationship between Near-Crashes and Crashes: Can Near-Crashes Serve as a Surrogate Safety Metric for Crashes? 2010. Available online: https://api.semanticscholar.org/CorpusID:6401904 (accessed on 11 February 2023).
- Yu, R.; Han, L.; Zhang, H. Trajectory data based freeway high-risk events prediction and its influencing factors analyses. Accid. Anal. Prev. 2021, 154, 106085. [Google Scholar] [CrossRef] [PubMed]
- Yuan, C.; Li, Y.; Huang, H.; Wang, S.; Sun, Z.; Li, Y. Using traffic flow characteristics to predict real-time conflict risk: A novel method for trajectory data analysis. Anal. Methods Accid. Res. 2022, 35, 100217. [Google Scholar] [CrossRef]
- Dingus, T.A.; Klauer, S.G.; Neale, V.L.; Petersen, A.; Lee, S.E.; Sudweeks, J.; Perez, M.A.; Hankey, J.; Ramsey, D.; Gupta, S.; et al. The 100-Car Naturalistic Driving Study, Phase II-Results of the 100-Car Field Experiment; Department of Transportation, National Highway Traffic Safety Administration: Washington, DC, USA, 2006.
- Allen, B.L.; Shin, B.T.; Cooper, P.J. Analysis of Traffic Conflicts and Collisions. Transp. Res. Rec. 1978, 67–74. [Google Scholar]
- Wang, C.; Xie, Y.; Huang, H.; Liu, P. A review of surrogate safety measures and their applications in connected and automated vehicles safety modeling. Accid. Anal. Prev. 2021, 157, 106157. [Google Scholar] [CrossRef] [PubMed]
- Cooper, D.F.; Ferguson, N. Traffic studies at T-Junctions. 2. A conflict simulation Record. Traffic Eng. Control. 1976, 17, 306–309. [Google Scholar]
- Shelby, S.G. Delta-V as a measure of traffic conflict severity. In Proceedings of the 3rd International Conference on Road Safety and Simulati, Indianapolis, IN, USA, 14–16 September 2011. [Google Scholar]
- Hayward, J.C. Near miss determination through use of a scale of danger. In Proceedings of the 51st Annual Meeting of the Highway Research Board, Washington, DC, USA, 17–21 January 1972. [Google Scholar]
- Ozbay, K.; Yang, H.; Bartin, B.; Mudigonda, S. Derivation and validation of new simulation-based surrogate safety measure. Transp. Res. Rec. 2008, 2083, 105–113. [Google Scholar] [CrossRef]
- Yang, H. Simulation-Based Evaluation of Traffic Safety Performance Using Surrogate Safety Measures. Ph.D. Thesis, Rutgers, The State University of New Jersey, Newark, NJ, USA, 2012. [Google Scholar]
- Pirdavani, A.; De Pauw, E.; Brijs, T.; Daniels, S.; Magis, M.; Bellemans, T.; Wets, G. Application of a rule-based approach in real-time crash risk prediction model development using loop detector data. Traffic Inj. Prev. 2015, 16, 786–791. [Google Scholar] [CrossRef]
- Bhatti, F.; Shah, M.A.; Maple, C. A novel internet of things-enabled accident detection and reporting system for smart city environments. Sensors 2019, 19, 2071. [Google Scholar] [CrossRef]
- Khan, A.; Bibi, F.; Dilshad, M.; Ahmed, S.; Ullah, Z.; Ali, H. Accident detection and smart rescue system using Android smartphone with real-time location tracking. Int. J. Adv. Comput. Sci. Appl. 2018, 9, 341–355. [Google Scholar] [CrossRef]
- Wang, L.; Abdel-Aty, M.; Shi, Q.; Park, J. Real-time crash prediction for expressway weaving segments. Transp. Res. Part C Emerg. Technol. 2015, 61, 1–10. [Google Scholar] [CrossRef]
- Yu, R.; Abdel-Aty, M.; Ahmed, M. Bayesian random effect models incorporating real-time weather and traffic data to investigate mountainous freeway hazardous factors. Accid. Anal. Prev. 2013, 50, 371–376. [Google Scholar] [CrossRef]
- Fu, C.; Sayed, T. Random parameters Bayesian hierarchical modeling of traffic conflict extremes for crash estimation. Accid. Anal. Prev. 2021, 157, 106159. [Google Scholar] [CrossRef]
- Hou, Q.; Tarko, A.P.; Meng, X. Analyzing crash frequency in freeway tunnels: A correlated random parameters approach. Accid. Anal. Prev. 2018, 111, 94–100. [Google Scholar] [CrossRef]
- Caliendo, C.; Guida, M.; Postiglione, F.; Russo, I. A Bayesian bivariate hierarchical model with correlated parameters for the analysis of road crashes in Italian tunnels. Stat. Methods Appl. 2022, 31, 109–131. [Google Scholar] [CrossRef]
- Yu, R.; Abdel-Aty, M. Analyzing crash injury severity for a mountainous freeway incorporating real-time traffic and weather data. Saf. Sci. 2014, 63, 50–56. [Google Scholar] [CrossRef]
- Jiang, H.; Deng, H. Traffic incident detection method based on factor analysis and weighted random forest. IEEE Access 2020, 8, 168394–168404. [Google Scholar] [CrossRef]
- Parsa, A.B.; Movahedi, A.; Taghipour, H.; Derrible, S.; Mohammadian, A.K. Toward safer highways, application of XGBoost and SHAP for real-time accident detection and feature analysis. Accid. Anal. Prev. 2020, 136, 105405. [Google Scholar] [CrossRef]
- Zhao, H.; Li, X.; Cheng, H.; Zhang, J.; Wang, Q.; Zhu, H. Deep learning-based prediction of traffic accidents risk for Internet of vehicles. China Commun. 2022, 19, 214–224. [Google Scholar] [CrossRef]
- Pawar, K.; Attar, V. Deep learning based detection and localization of road accidents from traffic surveillance videos. ICT Express 2022, 8, 379–387. [Google Scholar] [CrossRef]
- Karim, M.M.; Li, Y.; Qin, R.; Yin, Z. A dynamic spatial-temporal attention network for early anticipation of traffic accidents. IEEE Trans. Intell. Transp. Syst. 2022, 23, 9590–9600. [Google Scholar] [CrossRef]
- Yang, K.; Yu, R.; Wang, X.; Quddus, M.; Xue, L. How to determine an optimal threshold to classify real-time crash-prone traffic conditions? Accid. Anal. Prev. 2018, 117, 250–261. [Google Scholar] [CrossRef] [PubMed]
- Peng, Y.; Li, C.; Wang, K.; Gao, Z.; Yu, R. Examining imbalanced classification algorithms in predicting real-time traffic crash risk. Accid. Anal. Prev. 2020, 144, 105610. [Google Scholar] [CrossRef] [PubMed]
- Kurtc, V. Studyg car-following dynamics on the basis of the HighD dataset. Transp. Res. Rec. 2020, 2674, 813–822. [Google Scholar] [CrossRef]
- Schneider, P.; Butz, M.; Heinzemann, C.; Oehlerking, J.; Woehrle, M. Scenario-based threat metric evaluation based on the highd dataset. In Proceedings of the 2020 IEEE Intelligent Vehicles Symposium (IV), Las Vegas, NV, USA, 19 October–13 November 2020; pp. 213–218. [Google Scholar]
- Das, S.; Maurya, A.K. Defining time-to-collision thresholds by the type of lead vehicle in non-lane-based traffic environments. IEEE Trans. Intell. Transp. Syst. 2019, 21, 4972–4982. [Google Scholar] [CrossRef]
- Nadimi, N.; NaserAlavi, S.S.; Asadamraji, M. Calculating dynamic thresholds for critical time to collision as a safety measure. Proc. Inst. Civ. Eng.-Transp. 2022, 175, 403–412. [Google Scholar] [CrossRef]
- Jin, S.; Qu, X.; Wang, D. Assessment of expressway traffic safety using Gaussian mixture model based on time to collision. Int. J. Comput. Intell. Syst. 2011, 4, 1122–1130. [Google Scholar]
- Essa, M.; Sayed, T. Full Bayesian conflict-based models for real time safety evaluation of signalized intersections. Accid. Anal. Prev. 2019, 129, 367–381. [Google Scholar] [CrossRef] [PubMed]
- Yang, D.; Xie, K.; Ozbay, K.; Zhao, Z.; Yang, H. Copula-based joint modeling of crash count and conflict risk measures with accommodation of mixed count-continuous margins. Anal. Methods Accid. Res. 2021, 31, 100162. [Google Scholar] [CrossRef]
- Lee, C.; Saccomanno, F.; Hellinga, B. Analysis of crash precursors on instrumented freeways. Transp. Res. Rec. 2002, 1784, 1–8. [Google Scholar] [CrossRef]
- Yu, R.; Wang, X.; Yang, K.; Abdel-Aty, M. Crash risk analysis for Shanghai urban expressways: A Bayesian semi-parametric modeling approach. Accid. Anal. Prev. 2016, 95, 495–502. [Google Scholar] [CrossRef]
- Xiao, J. SVM and KNN ensemble learning for traffic incident detection. Phys. A Stat. Mech. Its Appl. 2019, 517, 29–35. [Google Scholar] [CrossRef]
- Qu, Y.; Lin, Z.; Li, H.; Zhang, X. Feature recognition of urban road traffic accidents based on GA-XGBoost in the context of big data. IEEE Access 2019, 7, 170106–170115. [Google Scholar] [CrossRef]
- Zhu, W.; Wu, J.; Fu, T.; Wang, J.; Zhang, J.; Shangguan, Q. Dynamic prediction of traffic incident duration on urban expressways: A deep learning approach based on LSTM and MLP. J. Intell. Connect. Veh. 2021, 4, 80–91. [Google Scholar] [CrossRef]
- Huang, J.; Ling, C.X. Using AUC and accuracy in evaluating learning algorithms. IEEE Trans. Knowl. Data Eng. 2005, 17, 299–310. [Google Scholar] [CrossRef]
- Sokolova, M.; Japkowicz, N.; Szpakowicz, S. Beyond accuracy, F-score and ROC: A family of discriminant measures for performance evaluation. In Proceedings of the Australasian Joint Conference on Artificial Intelligence, Hobart, Australia, 4–8 December 2006; pp. 1015–1021. [Google Scholar]
- Han, H.; Guo, X.; Yu, H. Variable selection using mean decrease accuracy and mean decrease gini based on random forest. In Proceedings of the 2016 7th IEEE International Conference on Software Engineering and Service Science (Icsess), Beijing, China, 26–28 August 2016; pp. 219–224. [Google Scholar]
- Parsa, A.B.; Taghipour, H.; Derrible, S.; Mohammadian, A.K. Real-time accident detection: Coping with imbalanced data. Accid. Anal. Prev. 2019, 129, 202–210. [Google Scholar] [CrossRef]
- Abou Elassad, Z.E.; Mousannif, H.; Al Moatassime, H. A proactive decision support system for predicting traffic crash events: A critical analysis of imbalanced class distribution. Knowl.-Based Syst. 2020, 205, 106314. [Google Scholar] [CrossRef]
- Chawla, N.V.; Bowyer, K.W.; Hall, L.O.; Kegelmeyer, W.P. SMOTE: Synthetic minority over-sampling technique. J. Artif. Intell. Res. 2002, 16, 321–357. [Google Scholar] [CrossRef]
- Elamrani Abou Elassad, Z.; Mousannif, H.; Al Moatassime, H. Class-imbalanced crash prediction based on real-time traffic and weather data: A driving simulator study. Traffic Inj. Prev. 2020, 21, 201–208. [Google Scholar] [CrossRef] [PubMed]
- Mehrannia, P.; Bagi, S.S.G.; Moshiri, B.; Al-Basir, O.A. Deep representation of imbalanced spatio-temporal traffic flow data for traffic accident detection. IET Intell. Transp. Syst. 2023, 17, 606–619. [Google Scholar] [CrossRef]
- Man, C.K.; Quddus, M.; Theofilatos, A. Transfer learning for spatio-temporal transferability of real-time crash prediction models. Accid. Anal. Prev. 2022, 165, 106511. [Google Scholar] [CrossRef]
- Zhang, Y.; Wang, H.; Zhang, D.; Wang, D. Deeprisk: A deep transfer learning approach to migratable traffic risk estimation in intelligent transportation using social sensing. In Proceedings of the 2019 15th International Conference on Distributed Computing in Sensor Systems (DCOSS), Santorini, Greece, 29–31 May 2019; pp. 123–130. [Google Scholar]
- Gürbüz, H.; Buyruk, S. Improvement of safe stopping distance and accident risk coefficient based on active driver sight field on real road conditions. IET Intell. Transp. Syst. 2019, 13, 1843–1850. [Google Scholar] [CrossRef]
LocationId | NumLanes | SpeedLimit | WeekDay | NumCars | NumTrucks | AvgSpeed |
---|---|---|---|---|---|---|
1 | 3 | Thu, Mon, Wed | 69,751 | 16,211 | ||
2 | 2 | Infinite speed | Tue | 2400 | 674 | |
3 | 3 | Thu | 2710 | 1037 | ||
4 | 3 | Infinite speed | Fri | 3799 | 952 | |
5 | 2 | Infinite speed | Fri | 8192 | 1887 | |
6 | 3 | Infinite speed | Wed | 2287 | 616 |
Category | Variable Level | Variable | Definition | Risk Event | Non-Risk Event | ||
---|---|---|---|---|---|---|---|
Mean | Standard Deviation | Mean | Standard Deviation | ||||
Traffic flow feature | Velocity | AvgV_U | Average upstream velocity | 25.906 | 7.959 | 30.081 | 4.722 |
AvgV_D | Average downstream velocity | 24.939 | 9.628 | 30.400 | 4.832 | ||
DiffV_UD | Difference of velocity between upstream and downstream | 3.496 | 3.053 | 2.428 | 1.912 | ||
StdV_U | Standard deviation of upstream velocity | 2.819 | 1.441 | 2.451 | 1.267 | ||
StdV_D | Standard deviation of downstream velocity | 2.568 | 1.400 | 2.515 | 1.322 | ||
CvV_U | Coefficient of variation of upstream velocity | 0.114 | 0.067 | 0.082 | 0.041 | ||
CvV_D | Coefficient of variation of downstream velocity | 0.128 | 0.121 | 0.083 | 0.043 | ||
Volume | Vo_U | Upstream volume (Veh/30 s) | 11.472 | 4.005 | 11.599 | 4.431 | |
Vo_D | Downstream volume (Veh/30 s) | 10.135 | 4.034 | 10.680 | 4.363 | ||
DiffVo_DU | Difference of volume between upstream and downstream (Veh/30 s) | 2.494 | 2.036 | 2.550 | 2.200 | ||
Difference | Diff_AvgV_U | Difference in average upstream velocity between main lane and adjacent lane | 4.567 | 2.668 | 5.034 | 2.426 | |
Diff_AvgV_D | Difference in average downstream velocity between main lane and adjacent lane | 4.464 | 2.919 | 4.995 | 2.481 | ||
Diff_StdV_U | Difference in standard deviation of upstream velocity between main lane and adjacent lane | 1.147 | 1.066 | 1.188 | 1.004 | ||
Diff_StdV_D | Difference in standard deviation of downstream velocity between main lane and adjacent lane | 1.166 | 1.106 | 1.244 | 1.058 | ||
Diff_CvV_U | Difference in coefficient of variation of upstream velocity between main lane and adjacent lane | 0.056 | 0.085 | 0.041 | 0.034 | ||
Diff_CvV_D | Difference in coefficient of variation of downstream velocity between main lane and adjacent lane | 0.063 | 0.088 | 0.043 | 0.036 | ||
Diff_Vo_U | Difference in upstream volume between main lane and adjacent lane (Veh/30 s) | 3.890 | 2.627 | 4.278 | 3.106 | ||
Diff_Vo_D | Difference in downstream volume between main lane and adjacent lane (Veh/30 s) | 3.684 | 2.697 | 4.117 | 2.947 | ||
Inter-vehicle kinematic feature | Velocity | Max_XV | Maximum longitudinal velocity | 29.226 | 9.309 | 32.936 | 5.247 |
Max_Diff_XV | Maximum difference of longitudinal velocity | 8.834 | 3.874 | 5.590 | 3.144 | ||
Max_YV | Maximum lateral velocity | 0.644 | 0.441 | 0.332 | 0.299 | ||
Acceleration | Max_XA | Maximum longitudinal acceleration | 1.395 | 1.026 | 0.689 | 0.445 | |
Max_Diff_XA | Maximum difference of longitudinal acceleration | 0.855 | 0.404 | 0.387 | 0.332 | ||
Max_YA | Maximum lateral acceleration | 0.312 | 0.185 | 0.153 | 0.103 | ||
Distance | Min_D | Minimum distance between vehicle | 17.137 | 8.080 | 39.897 | 40.379 |
True Risk Event | True Non-Risk Event | |
---|---|---|
Predicted risk event | True Positive (TP) | False Positive (FP) |
Predicted non-risk event | False Negative (FN) | True Negative (TF) |
Category | Risk Identification Model | Risk Prediction Model-5s | Risk Prediction Model-10s |
---|---|---|---|
Traffic flow feature | AvgV_D(0.043) | Diff_AvgV_D(0.085) | Diff_AvgV_D(0.080) |
Diff_AvgV_U(0.076) | Diff_AvgV_U(0.075) | ||
AvgV_D(0.060) | AvgV_D(0.065) | ||
Vo_U(0.051) | Vo_U(0.064) | ||
DiffVo_DU(0.059) | |||
Inter-vehicle kinematic feature | Min_D(0.198) | Max_XA(0.068) | Max_YA(0.046) |
Max_Diff_XA(0.177) | Max_Diff_XV(0.046) | ||
Max_YA(0.090) | |||
Max_XA(0.072) | |||
Max_Diff_XV(0.060) |
Original Dataset | SMOTE (Oversampling) | RENN (Undersampling) | ||||||
---|---|---|---|---|---|---|---|---|
Model | Metrics | Model | Metrics | Model | Metrics | |||
Accuracy | 0.982 | LR | Accuracy | 0.979 | LR | Accuracy | 0.982 | |
Precision | 0.521 | Precision | 0.460 | Precision | 0.524 | |||
Recall | 0.556 | Recall | 0.538 | Recall | 0.542 | |||
0.535 | 0.495 | F1 | 0.531 | |||||
0.967 | 0.968 | 0.967 | ||||||
Accuracy | 0.983 | Accuracy | 0.976 | Accuracy | 0.982 | |||
Precision | 0.624 | Precision | 0.389 | Precision | 0.552 | |||
Recall | 0.388 | Recall | 0.523 | Recall | 0.382 | |||
0.461 | 0.446 | 0.443 | ||||||
0.774 | 0.881 | 0.794 | ||||||
XGBoost | Accuracy | 0.986 | XGBoost | Accuracy | 0.986 | XGBoost | Accuracy | 0.986 |
Precision | 0.669 | Precision | 0.671 | Precision | 0.657 | |||
Recall | 0.539 | Recall | 0.534 | Recall | 0.561 | |||
F1 | 0.596 | F1 | 0.592 | 0.604 | ||||
0.975 | 0.978 | 0.976 | ||||||
Accuracy | 0.986 | Accuracy | 0.982 | Accuracy | 0.984 | |||
Precision | 0.673 | Precision | 0.528 | Precision | 0.586 | |||
Recall | 0.534 | Recall | 0.556 | Recall | 0.550 | |||
0.594 | 0.538 | 0.567 | ||||||
0.961 | 0.976 | 0.961 | ||||||
Accuracy | 0.983 | Accuracy | 0.983 | MLP | Accuracy | 0.983 | ||
Precision | 0.578 | Precision | 0.575 | Precision | 0.596 | |||
Recall | 0.491 | Recall | 0.430 | Recall | 0.515 | |||
0.521 | 0.489 | 0.536 | ||||||
0.957 | 0.963 | 0.962 |
Model | Accuracy | Precision | Recall | F1 | AUC |
---|---|---|---|---|---|
RF + Traffic flow features | 0.983 | 0.663 | 0.267 | 0.371 | 0.802 |
RF + Inter-vehicle kinematic features | 0.984 | 0.589 | 0.473 | 0.523 | 0.951 |
RF + Complete features | 0.986 | 0.673 | 0.534 | 0.594 | 0.961 |
XGBoost + Traffic flow features | 0.984 | 0.724 | 0.227 | 0.341 | 0.780 |
XGBoost + Inter-vehicle kinematic features | 0.982 | 0.537 | 0.505 | 0.519 | 0.965 |
XGBoost + Complete features | 0.986 | 0.669 | 0.539 | 0.596 | 0.975 |
Risk Identification Model | Risk Prediction Model-5s | Risk Prediction Model-10s | ||||||
---|---|---|---|---|---|---|---|---|
Model | Metrics | Model | Metrics | Model | Metrics | |||
LR | Accuracy | 0.982 | LR | Accuracy | 0.984 | LR | Accuracy | 0.983 |
Precision | 0.521 | Precision | 0.751 | Precision | 0.648 | |||
Recall | 0.556 | Recall | 0.212 | Recall | 0.220 | |||
0.535 | F1 | 0.329 | 0.326 | |||||
0.967 | 0.799 | 0.767 | ||||||
Accuracy | 0.983 | Accuracy | 0.983 | Accuracy | 0.983 | |||
Precision | 0.624 | Precision | 0.647 | Precision | 0.684 | |||
Recall | 0.388 | Recall | 0.223 | Recall | 0.237 | |||
0.461 | F1 | 0.330 | 0.347 | |||||
0.774 | 0.648 | 0.651 | ||||||
XGBoost | Accuracy | 0.986 | XGBoost | Accuracy | 0.983 | XGBoost | Accuracy | 0.983 |
Precision | 0.669 | Precision | 0.722 | Precision | 0.675 | |||
Recall | 0.539 | Recall | 0.243 | Recall | 0.249 | |||
F1 | 0.596 | 0.356 | 0.361 | |||||
0.975 | 0.801 | 0.778 | ||||||
Accuracy | 0.986 | Accuracy | 0.984 | Accuracy | 0.984 | |||
Precision | 0.673 | Precision | 0.749 | Precision | 0.720 | |||
Recall | 0.534 | Recall | 0.258 | Recall | 0.257 | |||
0.594 | 0.377 | 0.374 | ||||||
0.961 | 0.831 | 0.819 | ||||||
MLP | Accuracy | 0.983 | MLP | Accuracy | 0.982 | MLP | Accuracy | 0.983 |
Precision | 0.578 | Precision | 0.629 | Precision | 0.685 | |||
Recall | 0.491 | Recall | 0.217 | Recall | 0.221 | |||
0.521 | 0.316 | 0.331 | ||||||
0.957 | 0.743 | 0.729 |
Model | Accuracy | Precision | Recall | F1 | AUC | |
---|---|---|---|---|---|---|
Prediction Model-5s | RF + Traffic flow features | 0.984 | 0.734 | 0.247 | 0.367 | 0.820 |
RF + Inter-vehicle kinematic features | 0.984 | 0.764 | 0.222 | 0.343 | 0.715 | |
RF + Complete features | 0.984 | 0.749 | 0.258 | 0.377 | 0.831 | |
Prediction Model-10s | RF + Traffic flow features | 0.984 | 0.752 | 0.245 | 0.368 | 0.804 |
RF + Inter-vehicle kinematic features | 0.984 | 0.740 | 0.224 | 0.343 | 0.686 | |
RF + Complete features | 0.984 | 0.720 | 0.257 | 0.374 | 0.819 |
Authors | Feature Extraction | F1 | AUC | Sample Sized of Risk Event | Sample Sized of Non-Risk Event |
---|---|---|---|---|---|
Yu R et al. [17] | Kinematics characteristics of vehicle front 0∼5 s before risk occurrence | 0.866 | 0.960 | 256 | 1024 |
Yu et al. [17] | Kinematics characteristics of vehicle front 2∼5 s before risk occurrence | - | 0.970 | 256 | 1024 |
Yuan et al. [18] | Traffic flow characteristics of primary and secondary lanes 0∼30 s before risk occurrence | 0.447 | 0.871 | 129 | 3801 |
Katrakazas et al. [14] | Velocity, flow, and acceleration characteristics of polymerization 0∼300 s prior to risk occurrence | 0.335 | - | 3075 | 9225 |
This study | Traffic flow characteristics 0∼30 s before risk occurrence and kinematics characteristics between vehicles 0∼1 s before risk occurrence | 0.604 | 0.976 | 865 | 46,821 |
This study | Traffic flow characteristics of 5∼35 s before risk occurrence and kinematics characteristics of 5∼6 s between vehicles | 0.377 | 0.831 | 865 | 46,821 |
This study | Traffic flow characteristics 10∼40 s before risk occurrence and kinematics characteristics between vehicles 10∼11 s before risk occurrence | 0.374 | 0.819 | 865 | 46,821 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Huang, S.; Chen, H.; Wen, X.; Zhang, H. Predicting Highway Risk Event with Trajectory Data: A Joint Approach of Traffic Flow and Vehicle Kinematics. Electronics 2024, 13, 625. https://doi.org/10.3390/electronics13030625
Huang S, Chen H, Wen X, Zhang H. Predicting Highway Risk Event with Trajectory Data: A Joint Approach of Traffic Flow and Vehicle Kinematics. Electronics. 2024; 13(3):625. https://doi.org/10.3390/electronics13030625
Chicago/Turabian StyleHuang, Shichun, Haiyu Chen, Xin Wen, and Hui Zhang. 2024. "Predicting Highway Risk Event with Trajectory Data: A Joint Approach of Traffic Flow and Vehicle Kinematics" Electronics 13, no. 3: 625. https://doi.org/10.3390/electronics13030625
APA StyleHuang, S., Chen, H., Wen, X., & Zhang, H. (2024). Predicting Highway Risk Event with Trajectory Data: A Joint Approach of Traffic Flow and Vehicle Kinematics. Electronics, 13(3), 625. https://doi.org/10.3390/electronics13030625