A Recognition Method of Truck Drivers' Braking Patterns Based on FCM-LDA2vec

Area of Science:

• Transportation Science
• Data Mining
• Behavioral Analysis

Background:

• Truck driver behavior significantly impacts road safety and efficiency.
• Understanding braking patterns is crucial for accident prevention and traffic flow optimization.
• Existing methods for analyzing driver behavior lack specificity in pattern recognition.

Purpose of the Study:

• To propose and validate a novel method for recognizing truck drivers' braking patterns.
• To determine the distribution of different braking patterns during truck operation.
• To leverage vehicle data for driver behavior analysis and safety improvements.

Main Methods:

• Collected and segmented truck braking behavior data.
• Extracted 25 characteristic parameters and reduced dimensionality to seven key factors.
• Utilized Fuzzy C-Means (FCM) clustering and CH scores to identify nine braking behavior categories.
• Employed the LDA2vec model to identify three distinct truck driver braking patterns.

Main Results:

• Achieved an accuracy exceeding 85% in truck driver braking pattern recognition using the LDA2vec model.
• Successfully identified and categorized distinct braking patterns from operational vehicle data.
• Demonstrated the feasibility of mining driver behavior patterns from large datasets.

Conclusions:

• The developed LDA2vec-based model accurately recognizes truck drivers' braking patterns.
• Monitoring and targeted driver training based on identified patterns can prevent traffic accidents.
• This research contributes to enhancing road safety, protecting human health, and reducing environmental pollution.