Guizhou Province Freight Volume Forecast Research

Guizhou Province is located in the southwest mountain area,complex landform,surface rough crushing,shallow soil,serious soil erosion,environmental carrying capacity low,person much ground is little,ecological environment vulnerability,resulting in local traffic developed,restrict the healthy and rapid development of the economy,transportation system and economic system is in a state of disharmony.Accurately forecast the freight volume of Guizhou Province,to Guizhou Province transportation condition of reasonable allocation,improve the turnover rate of transport of goods,but also for Guizhou freight infrastructure planning proposals,rational allocation of resources,so that a more diversified modes of transport,between the various modes of transport in a more coordinated,promote the development of the transportation industry,which led to the development of the local economy.At present,the prediction of freight volume is mainly through two method,one method is to establish the forecast model based on the influence factors of freight volume,and the other method is to study the change of the time series data of freight volume.Each method has its advantages and disadvantages.Nowadays,many factors can influence the freight volume,but the forecast research only consider the properties of freight volume,the non characteristic parameters and complex unknown factors were not studied in those study.In this paper,the freight volume forecast of Guizhou Province is investigated by partially linear model based on the consideration of the above factors.The partially linear model is a model which contains some parameters and non parametric.At the same time,the multiple linear regression model and grey prediction model are used to forecast the freight volume of Guizhou Province,the prediction effect of the three kinds models is analyzed.The highway and railway freight volume forecast of Guizhou are study for the highway and railway freight way are the main freight way.Firstly,the freight volume influence factors of Guizhou are analysis.the main factors of influencing highway,railway freight volume is obtained by gray correlation analysis method.It is shown that the main influencing factors of the highway freight volume are area GDP,total retail sales of social consumer goods,fixed assets of estate investment,increase the total industrial value,output value of agricultural and sideline products,the total number of tourists,main influencing factors of railway freight volume are the total retail sales of social consumer goods,total industrial addedvalue,railway operating mileage,highway mileage,Maotai wine production,phosphate rock.Secondly,the partial linear,multiple regression and grey prediction forecast model are established by the main influencing factors mentioned above and the actual data between1990 to 2011.the fitting effect and prediction effect of those three models is analyzed by the actual data between 2012 to 2014.It is show that the prediction error of highway freight volume forecasting model less than 10%,the partial linear model and multiple regression model has better fitting effect and the partial linear model has the best prediction effect.For those three railway freight volume forecasting model,the fitting effect of the partial linear model is the best,the prediction error of railway freight volume by partial linear model less than 10%between 2012 to 2014,the prediction error by multiple regression and grey prediction forecast model more than 10%.Finally,the highway freight and railway freight volume of Guizhou Province between 2015 to 2017 are obtained by partially linear model.It is shown that the highway freight volume still maintain an upward trend,it will expected to exceed 10 million tons in 2017.the railway freight volume will started and restored to the original growth trend in 2016.The paper argues that the reasons for this result is partially linear model the parameter model and non parameter model together,the model not only focus on the excellent properties of model parameters,and avoid the non parameter model of the curse of dimensionality problem,so as to improve the model prediction accuracy.