Real-time highway accident prediction based on support vector machines

Traditional traffic accident prediction uses long-term traffic data such as annual average daily traffic and hourly volume. In contrast to traditional traffic accident prediction, real-time traffic accident prediction uses real-time traffic data, obtained from inductive loop detectors and usually collected every 20 or 30 seconds, to identify hazardous traffic conditions to potentially prevent the traffic accident occurrence. We aim at identifying traffic patterns leading to traffic accidents and not leading to traffic accidents in this study. Support vector machines (SVM) are used to classify traffic conditions into those two patterns with real-time traffic data. Traffic accident data and its corresponding real-time traffic data are collected from the traffic simulation software TSIS, which is a microscopic traffic simulation software. This is the first time the SVM method is applied for real-time traffic accident prediction. The experimental results show that it is promising for real-time traffic accident prediction by using the support vector machine method.