Related Experiment Video
Updated: Aug 23, 2026
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Identification of high-risk drivers using a classification and regression tree with experimental data from the driver
Sangjin Han1, Jinwoo Park1, Ja-Ho Leigh2,3
1Transport Studies Group, Department of Urban Planning, Graduate School of Environmental Studies, Seoul National University, Seoul, Korea.
Objective:
A driver's driving ability does not remain constant after a driver obtains a driving license. Over time, unexpected crashes, illnesses, and aging may impair physical and cognitive abilities, potentially resulting in a substantial decline in driving performance. However, in Korea, there are no quantitative standards for assessing driving aptitude based on physical and cognitive abilities. The purpose of this study is to propose criteria for assessing driving aptitude using experimental data from the Driver Test Station (DTS) and the Computerized Neurocognitive Test (CNT), which are widely used in the medical and psychological fields.
Methods:
A total of 175 participants underwent the DTS, CNT, and on-road driving assessment. After 10 disqualified participants were excluded, data from 165 participants were analyzed using the Classification and Regression Tree (CART) method to identify which specific measures in the DTS and CNT have the strongest impact on driving ability.
Results:
The T-score of the Trail Making Test Set A in the CNT was the most critical measure for assessing driving aptitude. Subsequently, reaction time-related measurements from the DTS tests, including the 6-Point Reaction Test, Emergency Brake Test, and Timer Test, were also proposed as complementary criteria for assessing driving aptitude. The model classified participants as passing or failing the on-road driving assessment, with accuracies of 0.70 and 0.60 on the training and validation data, respectively.
Conclusions:
The DTS and CNT data can be effectively used to identify drivers who are likely to fail the on-road driving assessment. The application of the developed tree model will help identify potentially high-risk drivers without an on-road driving assessment, thereby reducing the associated time and cost burden.
