In order to further explore the occurrence mechanism of road traffic accidents; this paper takes into account the differences in road types. Based on the open source traffic accident data set of the British Ministry of Transport, the traffic accident risk factors and their importance of the four road types with the highest traffic accident rates are studied. Firstly, based on data visualization, the distribution of data is preliminarily explored from the spatial and temporal dimensions. Secondly, aiming at the shortcomings of the integrated algorithm with more hyperparameters, which lacks accurate and efficient parameter adjustment methods, a Bayesian optimization based random forest algorithm (BO-RF) implemented by the Optuna framework is proposed to construct the risk factor importance level screening model of four road types. In addition, a control experiment with the random forest algorithm based on random search optimization (RS-RF) was carried out. The results show that the BO-RF algorithm model based on the Optuna hyperparameter optimization framework has shorter tuning time and higher accuracy. Finally, based on the visualization results of road traffic accident risk factor level obtained by the model constructed by BO-RF algorithm, a scheme suitable for reducing accidents under this road condition is proposed, which provides a reference for preventing road traffic accidents and optimizing relevant safety management regulations.
Study on Grade Discrimination Method of Traffic Accident Risk Factors Considering Road Type
Lect. Notes in Networks, Syst.
Intelligent Systems Conference ; 2023 ; Amsterdam, The Netherlands September 07, 2023 - September 08, 2023
2024-01-30
17 pages
Article/Chapter (Book)
Electronic Resource
English
Traffic accident discrimination method of intelligent vehicle road system
European Patent Office | 2023
|ESTIMATING ACCIDENT RISK LEVEL OF ROAD TRAFFIC PARTICIPANTS
European Patent Office | 2023
|ESTIMATING ACCIDENT RISK LEVEL OF ROAD TRAFFIC PARTICIPANTS
European Patent Office | 2022
|The road traffic accident prevention information-based accident risk notifying system
European Patent Office | 2023
|