- Journal
- ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
- Année
- 2025
- Volume
- 162
- Numéro
- C
- Article
- 112585
- Mois
- DEC 22
- DOI
- 10.1016/j.engappai.2025.112585
Abstract
Driving behavior is a critical factor in ensuring road safety, particularly in commercial transportation sectors where hiring safe and reliable drivers is a priority. This study presents a machine learning-based framework for predicting driver behavior using multimodal assessments, including psychological, physiological, and demographic factors. We utilized a driving simulator equipped with biometric sensors to capture the physiological data of the driver, including heart rate, eye blink rate, pupil diameter, and point of gaze (POG), during driving sessions. The psychological attributes are collected through a self-report questionnaire. The questionnaire is structured to obtain information about nine psychological characteristics of the driver, including instrumental attitude, social anxiety, sensation seeking, premeditation, urgency, selfishness, aggressive mode, life satisfaction, and conscientiousness. In addition, some demographic attributes, such as age and gender, are also adopted to study their effect on driving behavior. Experiments were conducted on 80 participants, each driving for 10 min. Various machine learning models, along with a feature selection strategy, were used to find the relationship between the driver's modalities and his driving behavior. Results demonstrate that the k-nearest neighbors (KNN) model achieved the best performance, yielding an accuracy of 93.75 % and a False Negative Rate (FNR) of 0. Feature importance analysis revealed that gaze distraction, sensation seeking, conscientiousness, and gender are the best predictors of driving behavior. The findings suggest that our model can serve as a valuable decision-support tool for taxi companies and transportation agencies aiming to enhance driver selection processes by identifying drivers with lower accident risks.