Combination prediction for short-term traffic flow based on artificial neural network

As the basis of urban traffic control and guidance, the prediction for short-term traffic flow is constrained by its dynamic properties. To build an optimum model and enhance the predicting accuracy of the traffic flow, a combination prediction algorithm based on neural network is proposed. According to the algorithm, the first Lyapunov exponent and recurrence plot are used to analyze the forecasting property of a traffic flow, and a set of predicting models are determined corresponding to the analysis. The predicted results of the traffic flow are obtained by a nonlinear combination model based on a neural network. Both simulated and real detected traffic volume are used to verify the effectiveness of the algorithm.