microsimulation

Machine Learning-Based Traffic Microsimulation Improves Crash Frequency Prediction, Study Finds

Researchers at the University of Leeds conducted microsimulation at five signalised intersections comparing a standard rule-based model and a machine learning behavior model. The ML model generated conflicts that led to crash predictions matching real-world data, while the rule-based model failed without calibration. The study highlights the potential of ML for proactive traffic safety assessment.