Near real time analytic processing of traffic data streams

Location data is vital for traffic management and for transportation and urban planning, but also benefits people in daily life, helping on decisions related to route planing and on the use of public transportation. Although historical data can provide insights on expected traffic volume at a certain region and time, predictions based solely on historical data fail to deal with events like street works and traffic-accidents. In this work, we use real time information together with historical data to predict traffic by road segment in the near future. The paper outlines the architecture of the system, the data model and the prediction method. Preliminary results using real world data on taxi positions show that using stochastic processes is a promising approach for short-term traffic forecasting.