Driving risk status prediction using Bayesian networks and logistic regression. Issue 7 (22nd September 2017)

• Record Type: Journal Article
• Authors: Yan, Lixin; Huang, Zhen; Zhang, Yishi; Zhang, Liyan; Zhu, Dunyao; Ran, Bin
• Is Part Of: IET intelligent transport systems. Volume 11:Issue 7(2017)
• Page Start: 431
• Page End: 439
• Publication Date: 2017-09-22
• Subjects: belief networks -- regression analysis -- traffic engineering computing; driving risk status prediction model -- Bayesian networks -- logistic regression algorithm -- receiver operating characteristic curve -- traffic accidents; Intelligent transportation systems -- Periodicals; Electronics in transportation -- Periodicals; 388.31205
• DOI: 10.1049/iet-its.2016.0207
• Languages: English
• ISSNs: 1751-956X

Abstract

Abstract : The ability to identify driving risk status plays an important role for reducing the number of traffic accidents. Bayesian networks (BNs) was applied to extract the main factors that significantly influence driving risk status. Five factors (driver state, sex, experience, vehicle state, and environment) were selected and considered to significantly influence driving risk status based on driving simulation experiments. Next, a logistic regression algorithm was employed to establish the driving risk status prediction model, and the receiver operating characteristic curve was adopted to evaluate the performance of the prediction model. The area under the curve was 0.903, indicating that the prediction model was both adaptable and practical. In addition, this study also compared three different models, namely modelling directly, modelling based on expert experience, and modelling based on BN. The results indicated that modelling based on BN outperformed all other methods. The conclusions could provide reference evidence for driver training and the development of danger warning products to significantly contribute to traffic safety.