Estimation of the Driving Style Based on the Users' Activity and Environment Influence

Area of Science:

• Human-Computer Interaction
• Transportation Engineering
• Data Science

Background:

• Driving style significantly impacts road safety and fuel efficiency.
• Existing methods for predicting driving style often rely on in-vehicle data collected during driving.
• There is a need for predictive models that can assess driving style based on pre-driving environmental and activity data.

Purpose of the Study:

• To develop and evaluate new models for predicting driving style based on user's environment and activity data.
• To assess the accuracy of predicting aggressive driving style using pre-driving parameters.
• To investigate the influence of novel parameters, such as car door operation, on prediction accuracy.

Main Methods:

• Collected 67 hours of driving data from 10 drivers.
• Developed an Android application (Sensoric) to gather low-level smartphone data on user activity.
• Incorporated new parameters like car door opening and closing manner into the prediction model.
• Compared predicted driving style against objective driving data for validation.

Main Results:

• Achieved encouraging results in predicting aggressive driving style.
• Precision values for aggressive driving recognition ranged from 0.727 to 0.909.
• Demonstrated that user's environment and activity data can predict aggressive driving style in advance.

Conclusions:

• User's environment and activity data are valuable predictors of aggressive driving style.
• The developed models and methods show potential for proactive driving behavior assessment.
• The inclusion of car door operation parameters improved prediction accuracy.