Research On Short - Term Traffic Flow Combination Forecasting Based On Neural Network And

Short-term traffic flow forecasting is a hot research field of intelligent transportation, short-term traffic flow prediction is one of the important issues of traffic control and vehicle navigation, and it is also one of the key technologies of intelligent traffic control and traffic-induced. Short-term traffic flow with nonlinear, time-varying, uncertainty, instability and other characteristics, in order to achieve a variety of short-term traffic flow forecasting model predicts that can ease urban traffic congestion and avoid waste of social resources. Therefore, the study of short-term traffic flow forecasting has important practical significance and application value.For traffic flow forecasting nonlinear time series, using the RBF neural network has a good effect, and its prediction accuracy is higher, but the RBF neural network has the disadvantage of poor generalization.In order to overcome the shortcomings of RBF neural networks, using support vector machines to predict traffic flow, but only a small SVM sample of200datas or less can get a better prediction results, but for general the traffic flow data samples is about one thousand, so using support vector machines for traffic flow forecasting also get a relatively large error.In order to overcome the RBF neural networks and support vector machines own shortcomings,use different combination algorithms to predict the different predictive results. The weight is calculated using the weights of the two prediction model calculation result obtained by the prediction error of two previous prediction result thereby achieving a new forecast, the new weight is worth to use the new more accurate predictions. This weighting formula is calculated based on empirical statistical results, but the prediction error is actually random and non-linear, so in fact, we can also take advantage of support vector machine to predict for the second time, so support vector machine to calculate the weight of two single forecasting model, resulting in a more accurate prediction. Experimental results show that the combination forecasting model is better than a single forecast. In contrast, the use of combination forecasting model to support forecasting results than using weight combination forecasting model calculation formula to calculate the weight vector machine to get the exact number, predict the effect will be better.