An evolutionary approach for short-term traffic flow forecasting service in intelligent transportation system

In recent years, traffic flow prediction has become a crucial technique in ITS (intelligent transportation system), which is helpful for alleviating the congestion in many metropolises and improving the efficiency of public traffic service. On the other hand, with the development of traffic sensors, traffic data are collected with a fantastic scale. It leads ITS into a data-driven application fashion. With this observation, it is a challenge to accurately and promptly forecast the traffic flow by effectively utilizing the big traffic data. In view of this challenge, in this paper, we propose an evolutionary method for short-term traffic flow forecasting service. Concretely, in our method, traffic flow is firstly specified by a model of time series. Then, the model is decomposed into seasonal component and the residual component. The seasonal component reflects the history average condition, while we treat the residual component as the output of a linear filter. The proposed method is evaluated with real bus transaction dataset. The experimental results show the effectiveness of our method.