Forecast of Passenger and Freight Traffic Volume Based on Elasticity Coefficient Method and Grey Model

Size: px
Start display at page:

Download "Forecast of Passenger and Freight Traffic Volume Based on Elasticity Coefficient Method and Grey Model"

Transcription

1 Available olie at ScieceDirect Procedia - Social ad Behavioral Scie ce s 96 ( 203 ) th COA Iteratioal Coferece of rasportatio Professioals (CICP 203) Forecast of Passeger ad Freight raffic Volume Based o Elasticity Coefficiet Method ad Grey Model Youa Wag a, Xumei Che a *, Yahui Ha a, Shuxia Guo b a MOE Key Laboratory for Urba rasportatio Complex Systems ad heory ad echology, School of raffic ad rasportatio, Beijig Jiaotog Uiversity, Beijig, 00044, Chia b Beijig Geeral Muicipal Egieerig Desig ad Research Istitute, Beijig, 00082, Chia Abstract he icrease i passeger ad freight traffic i a regio reflects the developmet of railways, highways, waterways, aviatio, ad pipelie. With the growth of ecoomy, Chia's trasportatio develops rapidly. However, the passeger ad freight traffic preset differet growth features i differet regios. herefore, a reasoable forecast model for passeger ad freight traffic ad the aalysis of relatioship betwee regioal trasportatio ad ecoomy are importat for trasportatio plaig. he elasticity coefficiet betwee the passeger traffic volume, freight traffic volume ad gross domestic product (GDP) is calculated based o the data from 200 to 200 i differet regios i Chia. he, the relatioship betwee the chage of regioal traffic volume ad regioal ecoomic developmet is obtaied. With the aalysis of the pros ad cos for differet forecast models, Elasticity Coefficiet Method, GM (, ) model, ad DGM model have bee used to forecast passeger ad freight traffic volumes from 20 to 205. I order to improve the accuracy of the forecast results, the combied models based o the variace reciprocal ad the optimal weightig are applied to optimize the forecastig model. Amog all the forecast models, the combied model with optimal weights outperforms other models with a relative error less tha 0.006% for the freight traffic volume. he accuracy of forecast models o passeger ad freight traffic volume has bee improved, which provides a reasoable basis for the plaig ad developmet of the trasportatio system. 203 he Authors. Published by Elsevier by Elsevier Ltd. Ope B.V. access uder CC BY-NC-ND licese. Selectio ad/or peer-review peer-review uder resposibility uder resposibility of Chiese of Overseas Chiese rasportatio Overseas rasportatio Associatio (COA). Associatio (COA). Key words Coefficiet of elasticity; passeger ad freight traffic volume; GM (, ) model; DGM model; combiatio model * Correspodig author. el.: ; fax: address: xmche@bjtu.edu.c he Authors. Published by Elsevier Ltd. Ope access uder CC BY-NC-ND licese. Selectio ad peer-review uder resposibility of Chiese Overseas rasportatio Associatio (COA). doi: 0.06/j.sbspro

2 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) Itroductio Forecastig o passeger ad freight traffic volume is the basis of plaig ad costructio of trasport facilities. At the same time, it provides valuable policy-makig iformatio for the govermet departmets which coducts market regulatio ad maagemet. Forecastig traffic volume accurately plays a importat role for the healthy developmet of trasportatio. Volume depeds o trasportatio demad, which is effected by regioal populatio, structure of ecoomy, idustrial scale ad layout, mechaizatio level, urbaizatio level ad social culture etc. herefore, how to forecast volume reasoably ad improve accuracy is a import issue. I this cotext, researchers pay much attetio to differet forecast approaches icludig qualitative forecast or quatitative forecast methods. Quatitative forecast methods iclude expoetial smoothig, grey forecastig model, ad regressio aalysis. I additio, Chia is a coutry with a vast territory. Regioal ecoomy developmet is ubalaced. Passeger ad freight traffic volume ad elasticity coefficiet of gross domestic product (GPD) differ greatly. herefore, the relatioship betwee traffic volume ad ecoomic developmet is difficult to aalyze. Aalyzig elasticity coefficiet is a effective approach to determie such relatioship. Meatime, the future passeger ad freight volume ca be obtaied by aalyzig elasticity coefficiet correctly. his paper forecast the passeger ad freight volume accordig to static ad dyamic elasticity coefficiet models. I order to improve accuracy, a grey model ad a combied model are also used to predict passeger ad freight volume of differet regios i Chia. 2. Passeger ad freight traffic volume forecast based o elasticity coefficiet method his paper has applied the elasticity coefficiet to forecast passeger ad freight volume firstly. Accordig to GPD data of differet regios from 200 to 200 ad the data of passeger ad freight traffic volume, elasticity coefficiet of passeger ad freight volume i differet regios ca be calculated. Further, elasticity coefficiet has bee used to forecast passeger ad freight traffic volume from 20 to 205 i differet regios. 2.. Elasticity coefficiet he trasportatio elasticity coefficiet is defied as a umerical measure of the relative respose of volume to chages i GDP, which ca be expressed as follows: rasportatio elasticity coefficiet = Rate of chage of volume / Rate of chage of GDP I this research, passeger ad freight traffic volume ad GDP i differet regios are used to calculate trasportatio elasticity coefficiet. Elasticity coefficiet ca be divided ito static ad dyamic elasticity coefficiet, depedig o differet methods of calculatio. he calculatio for static elasticity coefficiet is relatively simple ad used more frequetly Static elasticity coefficiet Curretly, the method used to determie trasportatio static elasticity coefficiet ca be divided ito two categories: oe is to calculate the value of elasticity coefficiet directly accordig to its defiitio, such as the geometric average method ad arithmetic average method; the other is to use regressio aalysis to determie the value of elasticity coefficiet. I this paper, a logarithmic liear regressio method is used to determie static elasticity coefficiet. Its equatio ca be expressed as: () Where is the traffic volume; A is coefficiet; E is trasportatio elasticity coefficiet; G is the GDP.

3 38 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) With the regioal data of passeger traffic volume, freight traffic volume ad GDP from 200 to 200, logarithmic liear regressios of differet regios was completed ad the static elasticity coefficiet of passeger ad freight traffic volume i differet regios was obtaied. he mea value of elasticity coefficiet for freight volume is , which meas the freight volume ad GDP is relatively ielastic. Elasticity coefficiets of Beijig ad Yua are ad , which reflects that the regioal freight volume does ot vary much with the regioal ecoomic growth. Elasticity coefficiets of Ahui ad ibet are.282 ad.4, which meas that the regioal ecoomic growth has effects o the freight volume growth. he mea value of elasticity coefficiet of passeger volume is , which meas lack of elasticity. However, compared with elasticity coefficiet of freight volume, the relatioship betwee passeger volume ad ecoomic developmet is closer. he elasticity coefficiets of passeger volume of Ier Mogolia, Heilogjiag, ad Yua are , , ad his result shows that the regioal passeger volume does ot icrease with the developmet of ecoomy. Elasticity coefficiets of Beijig ad iaji are.208 ad.466, which meas that the ecoomic developmet has facilitated the icrease of passeger volume Dyamic elasticity coefficiet I the model of dyamic elasticity coefficiet, elasticity coefficiet is calculated usig the relative chage rate betwee passeger, freight volume ad time. he regressio equatio o passeger volume ad time is fitted based o the data of from 200to 200. he regressio equatio betwee GDP ad time ca be derived. he, the elasticity coefficiet ca be calculated usig differetial equatio. Based o the followig Equatio (2), the elasticity coefficiet ca be calculated. he regressio equatios o passeger volume, freight volume, GDP, ad time t ca be obtaied usig quadratic curve fittig i MALAB. he, the derivatio of the regressio equatios is calculated ad iput to Equatio (2). Fially, the dyamic elasticity coefficiet for each regio is derived. For example, Elasticity coefficiets of freight/passeger volume from 20 to 205 of iaji are showed i able : able. Dyamic elasticity coefficiets of passeger ad freight volume i iaji (2) Year Passeger Freight Usig the dyamic elasticity coefficiet method, the time-varyig characteristics of elasticity coefficiet ca be cosidered. his method ca reflect the actual tred of trasportatio elasticity coefficiet ad the results caot be affected by the fluctuatios of the data. Because E,, G is a fuctio of time, the elasticity coefficiet i differet time ca be determied ad the future traffic volume ca be forecasted at the same time. Comparig with static elasticity coefficiet based method; this method does ot require that, G has a statistical relatioship. herefore, this method ca be applied i a wider rage Passeger ad freight volume forecast based o static elasticity coefficiet After elasticity coefficiet is determied, passeger ad freight volume i the future years ca be forecasted accordig to the models show as follow: (3) (4) Where is value of traffic volume at time t; is value of traffic volume at time ; E is coefficiet of elasticity; q is the average growth rate of ecoomy durig the ext period of time, %.

4 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) he growth rate of GDP i differet regio from 20 to 205 ca be obtaied from 2th Five-Year Pla. he, the static elasticity coefficiet ad the regioal ecoomic growth rates are iput to the Equatios (3) ad (4) i order to forecast the passeger ad freight volume of each regio from 20 to 205. For example, the results of Beijig are show i able 2: able 2. Forecastig results of passeger ad freight volume from 20 to 205 i Beijig based o static elasticity coefficiet Year Passeger(uit: 0000-people) Freight(uit: 0000-to) Passeger ad freight volume forecast based o dyamic elasticity coefficiet We iput the dyamic elasticity coefficiet ad the regioal GDP growth rates ito the Equatios (3) ad (4). he passeger ad freight volume of each regio from 20 to 205 ca be obtaied. For example, the results of Beijig are show i able 3: able 3. Forecastig results of passeger ad freight volume from 20 to 205 i Beijig based o dyamic elasticity coefficiet Year Passeger(uit: 0000-people) Freight(uit: 0000-to) Passeger ad freight traffic volume forecast based o grey model Regioal passeger ad freight traffic are affected by may factors. I order to forecast regioal passeger ad freight volume accurately, we must collect a great deal of iformatio ad cosider all kids of factors, which lead to the difficulty i establishig the forecast model. Eve the established model caot be used because of uavailable data. he grey system theory applies limited kow data to predict the behaviour of the ukow system. Grey predictio has caught much attetio of may researchers because of its high forecast accuracy, simple priciple, ad coveiet operatio. It has bee successfully applied i the passeger ad freight volume forecast. he paper uses GM (, ) model ad DGM model to forecast the regioal passeger ad freight volume from 20 to Passeger ad freight volume forecast based o GM (, ) model 3... GM (, ) model GM (, ) model is a importat grey theory based model, which is commoly used to forecast a system with limited data. he model is geerally used for short-term predictio. Origial form of the model is as follows: 0 X k ax k b (5) Differetial equatio ca be formulated as: dx ax b (6) dt Where a is the developmet coefficiet; b is grey volume.

5 40 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) We solve Equatio (6) ad obtai Equatio (7): b ak b x k x e (7) a a he solvig process of the model GM (, ) is as follows: ) Use regioal passeger /freight volume as the iitial data: X x, x 2, x (8) 2) Use accumulated method to obtai a ew sequece: X x, x 2, x (9) 3) Geerate earest sequece: Z k 0.5X k 0.5X k (0) 4) Set A a, b usig the least square method to solve A accordig to Equatio (): A B B B Y () WhereY x 2, x 3,, x, z 2 z 3 z B ; 5) Iput the obtaied a ito the equatio to obtai the value of Passeger ad freight volume forecast based o GM (, ) model x k ad save the series to forecast. Accordig to the GM (, ) model, passeger ad freight volume of differet regios from 20 to 205 has bee forecasted. For example, the forecastig results of Beijig are show i able 4: able 4. Forecastig results of passeger ad freight volume from 20 to 205 i Beijig based o GM (, ) model Year Passeger(uit: 0000-people) Freight(uit: 0000-to) Passeger ad freight volume forecast based o DGM model GM (, ) model is ofte used i the forecast for grey system. However, it is foud that whe the origial data sequece has a approximate expoetial growth patter, the forecast effects of GM (, ) model is ot good ad the forecastig results are ot stable. Whe the growth rate of traffic volume chages greatly, the forecast accuracy is low. Moreover, i the forecastig process, GM (, ) model depeds much o the iitial sequece ad a small chage i the iitial sequece may lead to large chage i the simulated sequece. o this ed, the discrete DGM model is itroduced i the forecast of passeger ad freight volume DGM model I the GM (, ) model, Equatio (5) is a discrete equatio, but Equatio (6) is a cotiuous equatio. I the forecast process, the parameter of Equatio (5) is iput to Equatio (6). Puttig parameter of discrete equatio ito cotiuous equatio is a modellig problem of GM (, ). herefore, DGM (, ) model is established. DGM (, ) gray differetial equatio is as follows:

6 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) x k x k (2) 2 Where, 2 If, 2 are parameters. is parameter colum ad x x 2 x 2 x 3 B, Y x x, he estimatio parameter colum of the least square parameter of the grey differetial Equatio (2) satisfies B B B Y (3) he estimated value of x k ca be expressed as: k k x k x 2 (4) Fially, the forecastig value ca be calculated usig Equatio (5) 0 k k x k x. (5) 2 he solvig process of the model DGM (, ) is as follows: ) Use regioal passeger/ freight volume as the iitial data: X x, x 2, x ; (6) 2) Use accumulated method to obtai a ew sequece: X x, x 2, x ; (7) 3) Calculate parameter, based o 2 B B B Y 4) Iput the obtaied, ito the Equatio (4) to obtai the value of () 2 x ( k ), ad the save the series 0 to forecast x k Passeger ad freight volume forecast based o DGM model Accordig to the DGM model, passeger ad freight volume of differet regios from 20 to 205are forecasted. For example, the forecastig results of Beijig are show i able 5: able 5. Forecastig results of passeger ad freight volume from 20 to 205 i Beijig based o the DGM model Year Passeger(uit: 0000-people) Freight(uit: 0000-to) Passeger ad freight traffic volume forecast based o a combied model I order to improve the forecast accuracy, a combied forecast model is established. he combied forecast method chooses appropriate weights for differet forecast methods. he forecastig result of each model multiplies by the weight ad sums up as the fial forecast results.

7 42 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) We set as forecast values of a combied model at time t ad as the forecast values of forecast model i at time t. he weight for forecast model i is. he combied model ca be expressed as follows: (8) With this model, passeger ad freight traffic volume ca be forecasted usig the reasoable weight. 4.. Determiatio of weight Whe usig the combied forecastig model, the weight is very importat. he reasoable weight ca improve the forecast accuracy sigificatly. Methods used to determie weights iclude the arithmetic mea method, the stadard deviatio method, the variace reciprocal method, the mea square reciprocal method, the AHP method, the Delphi method, ad the optimal weighted method. Amog them, the AHP method ad Delphi method are based o subjective assessmet, with errors. he arithmetic mea method is oe of the simple time series forecast method, which uses the average value of the observed time series durig a period of time as the ext forecast value. he variace reciprocal method gives higher weights to the model with lower square error. he optimal weighted method is based o a certai optimal criterio (such as the least squares criterio, mii or max criterio). For this method, the objective fuctio Q is built ad weight coefficiet of combied model is obtaied by miimizig Q uder the costraits (such as the sum of weight is). I order to improve the forecast accuracy, the paper has selected the variace reciprocal method ad optimal weighted method to determie the weight because these two methods have smaller error Variace reciprocal method he variace reciprocal method gives higher weights to the model with lower square error. he calculatio equatio is as follows: D i,,2,..., ; (9) i i i D i I Equatio (9), D i i it i ( ) 2 Where Di is the sum of square error of model i. Accordig to the website of Chia Idustry Research Report, the atioal freight volume is billio tos util December, 20 ad the growth rate is 3.7% compared with the same period of last year. he passeger volume reaches 35.7 billio people, with a icrease of 7.6% compared with the same period of last year. he paper applies the static elasticity coefficiet model, the dyamic elasticity coefficiet model, the GM (, ) model, ad the DGM model to forecast the 20 atioal freight volumes, which are billio tos, billio tos, billio tos, ad billio tos. he relative errors are , , 0.035, ad he forecastig results with the static elastic coefficiet model, dyamic elastic coefficiet model, GM (, ) model, ad DGM model are show i able 6. As show i able 6, the relative errors vary from the models. I order to optimize the forecast results, the variace reciprocal method is used to determie weight. he forecastig results of atioal passeger ad freight volume of 20 are iput to the Equatios (9), (20), ad (2). he obtaied weight of static, dyamic elasticity coefficiet model, GM (, ) model, ad DGM model are show i able 6: (20) (2)

8 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) able 6. Weights of differet forecast methods for passeger ad freight volume Passeger Freight GM(,) Static elasticity coefficiet model Dyamic elasticity coefficiet model DGM Volume (uit:0000-people) Relative error (%) weight Volume (uit:0000-to) Relative error (%) weight Actual value of 20 freight traffic Optimal weighted method Optimal weighted method is i accordace with the priciple of izig the square error of the combied model durig a past obtai weight of each forecast model. he forecast error of model i at time t is as follows: eit yi y it ; i, 2,, (22) he forecast error vector of model i ca be expressed as: F [, 2,..., ] i ei ei e i (23) Error matrix is as follows: e [ F, F,..., F ] (24) 2 Error iformatio matrix E r e e E E Er is expressed as: E E Set R [,,...,] r as -dimesioal vector i which all elemets are ; Set W [, 2,... ] as - dimesioal vector of the weight for forecast model. he sum square of forecast error of the combied model S is: m m 2 2 t ( i it) r t t i S e e W E W (26) Accordig to a liear programmig show as mi S W ErW the optimal solutio is as follows: st RW, Er Rr W (27) R r E r R r he optimum weights of each forecast model are calculated usig MALAB, the weight coefficiets are show i able 7: able 7. Weights of differet forecast methods usig optimal weight GM(,) Static elasticity coefficiet model Dyamic elasticity coefficiet model DGM Passeger weight Freight weight (25)

9 44 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) Freight volume forecast based o the combied model of he combied model forecast based o variace reciprocal method Accordig to the idetified weight of the static elasticity coefficiet method, dyamic elastic coefficiet method, ad grey model forecast method, the passeger ad freight volume of 20 is multiplied by the relevat weight. For example, the forecastig results of Beijig ad Chia usig the combied model with optimum weight are show i able 8: able 8. Forecastig results of passeger ad freight volume i 20 i Beijig ad i Chia based o the variace reciprocal method Passeger (uit: 0000-people) Freight (uit: 0000-to) Regio GM(,) model Static elasticity coefficiet model Dyamic elasticity coefficiet model DGM model Combied model Beijig Natioal Beijig Natioal Usig the combied model based o variace reciprocal method, the 20 freight volume of Chia is forecast as billio tos ad the relative error is.33 %.Compared with the DGM model, the relative error icreases. he forecast passeger volume of 20 i Chia is billio people ad the relative error is 0.3 %. Accordig to the results, the relative error is reduced compared with a sigle forecast model. hus, the forecast accuracy is improved he combied model forecast based o optimal weighted method Accordig to the idetified weight of the static elasticity coefficiet forecast method, dyamic elastic coefficiet method, ad grey model forecast method, the passeger ad freight volume of 20 is multiplied by the relevat weight. he forecastig results ca be obtaied by summig the products. For example, the forecastig results of the combied model with optimum weight are show i able 9: able 9. Forecastig results of passeger ad freight volume i 20 i Beijig ad i Chia based o the optimal weighted method Passeger (uit: 0000-people) Freight (uit: 0000-to) Regio GM(,) Static elasticity Dyamic elasticity DGM Combied model coefficiet model coefficiet model model model Beijig Natioal Beijig Natioal Based o the combied model with optimum weight, the atioal freight volume of 20 is forecasted as billio tos. he relative error is reduced to 0.0 %, which is lower tha a sigle model. he forecastig atioal passeger volume of 20 is billio people ad the relative error is reduced to %. Compared with the forecastig results of a sigle model, the forecastig accuracy has bee greatly improved. I coclusio, whether the combied model based o the variace reciprocal method or optimal weighted method is able to reduce forecast error. he combied model with optimum weight improves the accuracy of combied forecast model more effectively. However, the variace reciprocal method is simpler tha the optimal weighted method. I practice, how to select these methods depeds o real situatio.

10 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) Passeger ad freight volume forecast based o optimal weighted method from 202 to 205 Accordig to the comparative aalysis i the sectio of 4.2, the research has selected the combied model with optimum weight to forecast the passeger ad freight traffic volume from 202 to 205. It is impossible to determie optimum weight of each model from 202 to 205 because passeger ad freight traffic volume from 202 to 205 is ukow. he paper applies the obtaied optimum weight of a sigle model based o the data of 20 to forecast the passeger ad freight volume from 202 to 205. he results are show i able0: able 0. Forecastig results of freight volume from 202 to 205 based o the optimal weighted method Passeger (uit: 0000-people) Freight (uit: 0000-to) Beijig iaji Hebei Shaxi Ier Mogolia Liaoig Jili Heilogjiag Shaghai Jiagsu Zhejiag Ahui Fujia Jiagxi Shadog Hea Hubei Hua Guagdog Guagxi Haia Chogqig Sichua Guizhou Yua ibet Shaaxi Gasu Qighai

11 46 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) Nigxia Xijiag Not Classified Natioal able 0 idicates that the atioal freight volume growth rate remais at o more tha 5% ad differet regios show differet growth rates. Most regios maitai a average aual growth rate higher tha 0%, such as Beijig, Ier Mogolia, Shaxi, Heilogjiag, ad Zhejiag. However, average aual growth rate of ibet, Nigxia ad other wester provices are higher tha 20%. With the rapid developmet of the wester regio ad cotiuous improvemet of the trasportatio system i wester regio, freight volume will icrease more sigificatly. I additio, the passeger volume also icreases with a average aual growth rate at o more tha 5%. I most regios, passeger volume growth rates are at the same level ad icrease ot sigificatly. However, the growth rate is higher i Beijig, iaji, Ier Mogolia, Xijiag ad ibet. he large populatio ad rapid ecoomic growth of Beijig ad iaji are the mai reasos why the passeger traffic volume icreases. he passeger volume growth rates of ibet ad Xijiag are much higher tha the atioal average value. he reaso for this is that the rapid ecoomic ad trasportatio developmet of wester regios i recet years. Especially the costructio of Qighai-ibet lie results i rapid growth of the passeger volume. Furthermore, the adjacet regios or the regios with similar ecoomic coditios preset the similar characteristics, such as the Beijig-iaji-agsha ecoomic circle. he elasticity coefficiet of passeger ad freight volume i Beijig ad iaji is close. Besides, the freight volume elasticity coefficiets of Hea ad Hebei, which are adjacet regios, are close. I Qighai, Nigxia, ad Xijiag, which are adjacet regios, the relatioship betwee passeger/freight volume ad ecoomic developmet are similar.. 5. Coclusios he method for passeger ad freight volume forecast has focused o quatitative ad qualitative forecast. I this paper, we use elasticity coefficiet, GM (, ), DGM model ad combied model to forecast passeger ad freight volume. Some coclusios ca be obtaied as follows: ) he elasticity coefficiet of passeger ad freight volume are calculated ad aalysed based o the regioal passegers ad freight volume from 200 to 200, ad GDP data. he differet relatioships ad characteristics betwee the ecoomic developmet ad volume growth of differet regios ca be foud. 2) he passeger ad freight volume of differet regios i Chia from 20 to 205 are forecasted based o static ad dyamic elasticity coefficiet model. 3) Grey model ca elimiate the fluctuatio of the freight/passeger volume sequece effectively. he passeger ad freight traffic volume from 20 to 205 i each regio i Chia is forecasted based o the GM (, ad DGM model. 4) I compariso with the actual passeger ad freight volume i 20, the accuracy of the established forecast model was tested. he combied model with optimal weights outperforms other precisio models with a relative error of the freight traffic volume less tha 0.006%. Although the forecast accuracy of passeger ad freight volume is improved i this research, which provides a reasoable basis for the plaig ad developmet of the trasportatio system, we still eed to refie the relatioship betwee ecoomic developmet ad the icrease of passeger ad freight volume i the future.

12 Youa Wag et al. / Procedia - Social ad Behavioral Scieces 96 ( 203 ) Ackowledgemets his research is supported by Program for New Cetury Excellet alets i Uiversity: NCE Refereces Du, J. (2009). Research o determiatio of elasticity coefficiet o highway trasportatio volume predictio. Beijig: Beijig Uiversity of echology. Li, X. G. (2009). Research o calculatio of elasticity coefficiet o highway trasport based o the trasport ad ecoomic dyamic aalysis of highway. Beijig: Beijig Jiaotog Uiversity. Zhu, Y. Q., Liao, X. Q., & Zhu H. Y. (20). Chiese railway freight volume combied forecastig. Moder Busiess, 7, Yag, Z. S., & Wag, D. H. (994). Determiatio ad applicatio of trasportatio elasticity coefficiet. Joural of Jili Uiversity of echology, 3, Qiao, X. M., Dog,, M, & Zhag, M. X. (2004). Research o highway passeger ad freight traffic volume based o the elasticity coefficiet. East Chia Highway, 5, Liu, S. F., & Dag Y. G. (2005). Predictio theory ad methods. Beijig: Higher Educatio Press. Jiao, Y. L., & Su, B. Z. (2008). Predictio of railway passeger ad freight traffic based o gray theory. Joural of Lazhou Jiaotog Uiversity, 27, Natioal Bureau of Statistics of Chia. ( ). Chia Statistical Yearbook. Chia Statistics Press. Ha, Y. H., & Che, C. (20). Relatioship betwee freight trasportatio ad ecoomic growth of Chia. Joural of Beijig Istitute of echology, 20, ruog, D. Q., & Ah, K. K. (202). A accurate sigal estimator usig a ovel smart adaptive grey model SAGM(,). Expert Systems with Applicatios, 39,