Adopting Machine Learning and Spatial Analysis Techniques for Driver Risk Assessment: Insights from a Case Study.

Area of Science:

• Traffic Safety Engineering
• Geographic Information Systems (GIS)
• Machine Learning (ML)

Background:

• Traffic violations, often stemming from aggressive driving, are a major cause of road crashes.
• Driver actions, intentional or not, endanger lives and property.
• Understanding violation patterns is crucial for effective traffic management.

Purpose of the Study:

• To investigate traffic violations using spatial analysis and machine learning.
• To identify high-risk violation zones on expressways.
• To classify and predict different types of traffic violations.

Main Methods:

• Georeferenced traffic violation data from Luzhou, China (2016).
• Spatial analysis with Inverse Distance Weighted (IDW) interpolation for hotspot mapping.
• Machine learning models: K Nearest Neighbors (KNN), Support Vector Machine (SVM), CN2 Rule Inducer.

Main Results:

• Wrong-way driving was the most frequent violation.
• IDW identified violation hotspots for targeted interventions.
• KNN (k=7, Manhattan distance) achieved 99% accuracy, outperforming SVM and CN2.
• Multiple evaluation metrics (AUC, F-score, precision, recall, specificity) assessed model performance.

Conclusions:

• Machine learning, particularly KNN, effectively predicts traffic violations.
• Spatial analysis aids in resource allocation for high-risk areas.
• Findings offer insights for engineering and traffic control measures to enhance road safety.