Traffic Flow Forecasting Method Based On Combination Model

In the new century, traffic problem has been concerned more and more by many countries. Traffic congestion, environmental pollution and traffic accident has influenced the development of society and people's life seriously. Intelligent Transportation System (ITS) is the solution of these problems. More and more Scientists pay attention to the research of ITS. Traffic Flow Guidance System (TFGS) takes important part in ITS. Traffic flow prediction is the core issue in the TFGS. How to predict traffic flow amount online is the key problem to the TFGS. The prediction accuracy is most important in TFGS. In this thesis, we are researching a traffic flow forecasting model and validating the model.In this thesis, based on the analysis of the properties of traffic flows, we find the causes of the exceptional data and propose corresponding solutions as categories. After the handle of the traffic data, which is the real traffic conditions, it can be the input of the traffic flow forecasting model, and improve the prediction accuracy. Based on the analysis of other models, we adopt a new traffic flow forecasting model based on combination model, which has higher prediction accuracy. The prediction accuracy can meet the actual requirements. The main contributions are listed as follows.(1) Proposing different solutions of different categories of traffic error data, and check the traffic data by three methods. It can make sure that the error data does not impact the prediction accuracy.(2) Create three traffic flow forecasting models:Elman Neural Network, Historical Trend, Non-parametric Regression. Introduce the implementation and principle of the three models.(3) Proposing the combination model, create a traffic flow forecasting model based on the other three models. Validate the model through the forecasting of the sample data.