Using machine learning techniques to characterize sleep-deprived driving behavior

Area of Science:

• Neuroscience
• Traffic Safety
• Machine Learning

Background:

• Sleep deprivation is a significant risk factor for driving impairment and accidents.
• Previous research utilized machine learning to characterize driving behavior under the influence of alcohol and alprazolam.

Purpose of the Study:

• To classify abnormal driving behavior specifically induced by sleep deprivation.
• To evaluate if a machine learning model developed for sleep deprivation can also identify driving impairments from other interventions.

Main Methods:

• A gradient boosting machine learning model was employed to classify driving behavior.
• Data from 24 subjects tested under sleep-deprived and well-rested conditions were analyzed.
• The model was validated using 5-fold cross-validation and tested against data from alprazolam, alcohol, and placebo conditions.

Main Results:

• The sleep deprivation model achieved an accuracy of 77% ± 9% in detecting abnormal driving behavior in a simulator.
• Driving behavior after alprazolam and, to a lesser extent, alcohol intake exhibited characteristics similar to those observed during sleep deprivation.
• Probability scores indicated significant overlap between sleep deprivation and alprazolam/alcohol impairment, with placebo showing minimal overlap.

Conclusions:

• A machine learning model effectively detects driving impairments caused by sleep deprivation.
• The model highlights shared driving characteristics between sleep deprivation and other impairing substances like alcohol and alprazolam.
• This model can serve as a benchmark for evaluating the driving safety impact of new pharmaceutical interventions.