Short-time Traffic Flow Volume Prediction Based on Support Vector Machine with Time-dependent Structure

摘要

Using support vector machine (SVM) with a time-dependent structure, a new model is proposed to predict short-time traffic flow volume. In order to match the time varying characteristic of the traffic flow volume, in the developed model, each prediction requires a reconstruction process of SVM structure. The current SVM structure is determined by restraining with the input of the data of the traffic flow volume in the last hour. Then the predicted value is obtained according to the current SVM structure. The experimental results show that the prediction model with a time-dependent structure SVM outperforms the one without a time-dependent structure. Especially during the period from 7:00 a.m. to 22:00 p.m., the absolute mean error and mean squared error of the prediction model are 5.1veh/5min, 6.0veh/5min, respectively.