The Urban Arterial Traffic Flow Forecasting Based on BP Neural Network

Abstract

Predictive analytics of the traffic flow is paid more attention by the traffic engineering experts and relevant departments. However, how to forecast traffic volume still is an important problem affecting the traffic theoretical and practical analysis. Firstly, this paper set up a three layers BP neural network basing on the actual situation to introduce the modeling process of the neural network in detail, and forecast the short-term traffic volume by the means of rolling forecast. Secondly, taking the Hailar Street in Hohhot for example, two groups of test data from the same time of different days and sequent time of the same day were trained and forecast. in addition, predicting results and actual results were compared, and the correlations between test data and predicting result was analyzed and disclosed. Finally, the conclusion shows the error is acceptable and BP Neural Network constructed is practical when prediction accuracy is not very high.

An effective co-evolutionary algorithm based on artificial bee colony and differential evolution for time series predicting optimization

“…Parametric methods can obtain a good performance when traffic flow shows regular variations, but they cannot deal with the obvious chaotic characteristics and nonlinear nature of traffic flow data. Therefore, many researchers have focused on nonparametric methods, such as K-nearest neighbor (KNN) model [6], back propagation (BP) neural network model [7], radial basis function (RBF) neural network model [8], and support vector regression (SVR) [9]. In these classical models, RBF neural networks have not only more powerful approximation but also better autoadaptability.…”