Short-Term Traffic Flow Prediction Based on Upstream GA-MLR Prediction and Coefficients of Links

Traffic detectors can obtain many kinds of traffic information, which is helpful to decision-makers. However, these detectors are easily damaged or their data is lost during transmission. Without enough traffic flow data, the accurate traffic flow and its trend cannot easily be found. This paper presents a method for short-term traffic flow forecasting based on the upstream link traffic flow predicted and the network coefficients of upstream link and target links. Based on a large number of intersection data, the coefficients between upstream link and downstream one are calculated and GA-MLR prediction model of upstream link is trained. The traffic flow of the downstream link can be got by using the predicted traffic flow of a single upstream link and the network coefficients between links. Finally, the experiment prediction results show that the prediction method is effective for the downstream traffic flow prediction.