Freight Volume Forecast (FVF) plays an important role in national and regionaldevelopment plan, as well as a base of investment decision making for transportinfrastructure. Research on FVF aims at how to mine useful information fromtransportation system and use forecast model to process it, in order to provideaccurate and effective FVF for traffic management department and transportationenterprise to help them arrange transportation reasonably and control transportationprocess conveniently.Following research on FVF is completed in this paper after analyzing relevantstudies at home and abroad.Firstly, the paper point out that calculation accuracy is not high adopting theRBFNN (Radial Basis Function Neural Network) model to forecast time series datawith growth trends.Secondly, a new mixed RBFNN is projected, which was composed of a RBFNNand a linear regression model time-related. The new model is particularly suitable forforecasting time series data with growth trends.Thirdly, a hybrid learning algorithm for RBFNN based on nearestneighbor-clustering algorithm centers-selected and gradient descent training isproposed to improve calculation accuracy and training speed. And the paper writescode by using matlab programming language. A Cargo Forecasts example of SuzhouCity confirmed the effectiveness and practicality of the new model and algorithm.Lastly, the paper presents Auto Regressive Integrated Moving Average (ARIMA)models. Employing the ARIMA and RBFNN as prophase forecasting model, thispaper projected an optimal combination forecasting model. The efficiency of themodel was illustrated through forecasting the freight volume of Suzhou City. |