From Anger Detection to Intensity Modeling: A Two-Stage Machine Learning Approach Using Driving Performance and...

Area of Science:

• Occupational Safety and Health
• Machine Learning Applications
• Transportation Psychology

Background:

• Driver anger is a significant safety concern in occupational settings, particularly for commercial drivers.
• Existing methods for detecting driver anger often lack accuracy and fail to differentiate intensity levels.
• Frequent exposure to anger-eliciting situations impacts professional drivers' well-being and job performance.

Purpose of the Study:

• To develop and validate a machine learning framework for modeling driver anger intensity.
• To improve the accuracy of anger detection and reduce misclassification of neutral states.
• To explore the occupational applications of driver anger intensity detection for enhanced safety and worker well-being.

Main Methods:

• A two-stage machine learning framework was employed, first detecting anger presence, then classifying its intensity.
• Combined driving performance metrics and physiological signals were utilized as input features.
• Performance was compared against a single-stage four-class model.

Main Results:

• The two-stage model significantly improved accuracy in detecting and classifying driver anger intensity.
• A substantial reduction in misclassification of neutral emotional states was observed.
• Feasible modeling of driver anger across three intensity levels was demonstrated.

Conclusions:

• Driver anger intensity can be effectively modeled using a combination of driving data and physiological signals.
• The proposed two-stage machine learning framework offers a more accurate approach to anger detection in occupational drivers.
• Findings support the integration of anger-intensity detection into driver monitoring systems for adaptive assistance and operational improvements, enhancing fleet safety and driver well-being.