Real-time traffic event detection using Twitter data

Incident detection is an important component of intelligent transport systems and plays a key role in urban traffic management and provision of traveller information services. Due to its importance, a wide number of researchers have developed different algorithms for real-time incident detection. However, the main limitation of existing techniques is that they do not work well in conditions where random factors could influence traffic flows. Twitter is a valuable source of information as its users post events as they happen or shortly after. Therefore, Twitter data have been used to predict a wide variety of real-time outcomes. This paper aims to present a methodology for a real-time traffic event detection using Twitter. Tweets are obtained through the Twitter streaming application programming interface in real time with a geolocation filter. Then, the author used natural language processing techniques to process the tweets before they are fed into a text classification algorithm that identifies if it is traffic related or not. The authors implemented their methodology in the West Midlands region in the UK and obtained an overall accuracy of 92.86%.