Modeling Urban Freeway Rear-End Collision Risk Using Machine Learning Algorithms

A large amount of traffic crash investigations have shown that rear-end collisions are the main type on freeway. The purpose this study is to investigate collision risk Firstly, a new framework was proposed develop probability (RCP) model between two vehicles based Generalized Pareto Distribution (GPD). Secondly, freeway (F-RCR) defined as sum each vehicle and divided into three levels which high, median, low risk. Then, different machine learning algorithms were used F-RCR under condition an unbalanced dataset. result RCP showed continuous change can identify dangerous quickly compared traditional models even when speed leading faster than following vehicle. When distribution road difference adjacent lanes volume large, will increase. Multi-Layer Perceptron (MLP) found be more suitable for modeling F-RCR. provided in research transferrable proactive safety management system.