Research On Short-term Tourist Flow Forecasting Combination Model Of Mountain Scenic Spot

In recent years,with the development of national economy and the increase of residents’ income,the tourism industry has become one of the fastest growing industries.The tourist boom not only brings huge economic benefits to countries and regions,but also bring challengs to the daily management and services of the scenic spots,especially to the mountain scenic spot,its complex geographical characteristics makes the daily management and services more difficult,so the short-term tourist flow forecasting of mountain scenic spot is crucially improtant.However,due to the influence of various factors such as peak and off season,weather,vacation and event activities,the short-term tourist flow of mountain scenic spot in China shows strong non-linear characteristic,and the single forecasting models that always have drawbacks are difficult to achieve stable prediction.The combination model makes the forecasting more robust by combining the advantages and avoiding shortcomings of single models,so it enables forecasting the short-term tourist flow in the mountain scenic spot accurately,and then it provides a direct,scientific and effective reference for the management decision of the scenic spot,and even the entire tourism industry.Huangshan Scenic Spot,which has all the characteristics of mountain scenic spot,is famous in China.To study the short-term tourist flow forecasting of mountain scenic spot,based on the projects that undertaken by my research group,this dissertation takes Huangshan Scenic Spot as the research case to study the weekly tourist flow forecasting for the management demand of scenic spot.Meanwhile,according to the characteristics of tourist flow,the daily tourist flow forecasting was divided into two types:peak season and off season daily tourist flow forecasting which were studied respectively.The three types of short-term tourist flow are forecasted by several combination models which pairwised combine SVR,BP,ARMA and RF four single models.The main contents of this dissertation are as follows:(1)The dissertation analyzes the distribution characteristics of annual,monthly,weekly and daily tourist flow in mountain scenic spot,and divides daily tourist flow into peak season daily and off season daily tourist flow based on distribution characteristics and monthly tourist flow ratio.The related forecasting variables of short-term tourist flow are also analyzed according to the daily tourist flow and other relevant data.The distribution characteristics and related forecasting variables provide the gists for the classification forecasting of short-term tourist flow.(2)The weekly tourist flow forecasting model of scenic spots is established.Due to the strong non-linear and weak linear characteristics of the weekly tourist flow,the forecasting combination models SVR-ARMA or BP-ARMA is established.The two models forecast the weekly tourist flow with SVR or BP firstly,and then use the ARMA to correct the SVR’s or BP’s forecasting error.Subsequently,the forecasting value of SVR or BP is added to the correction values to obtain the final prediction results.In the empirical process,the relevant quantitative input variables are determined by the correlation method,and a dynamic dummy variable for model input is proposed on the two qualitative variables of statutory holiday and weather.Finally,the prediction results comparison show that SVR-ARMA or BP-ARMA is better than SVR or BP,and SVR-ARMA is more suitable than BP-ARMA for weekly tourist flow forecasting.(3)The peak season daily tourist flow forecasting model of scenic spots is established.Considering that the goodness of fit,a GFW combination model is proposed for the daily tourist flow of peak season which only has the non-linear characteristic.The model first uses three single nonlinear models SVR,BP and RF to forecast,and then obtain the final prediction result by pairwise combining the forecasting values of the three single models base on GFW and three combination models LCM,GEOM-WTD and HARM-WTD.According to the daily tourist flow of peak season and relevant data in scenic spot from 2011 to 2015,an empirical study is developed.The result shows that the GFW combination model is superior to the single models and the prediction result is more robust.Meanwhile,the prediction results’ comparison between the GFW combination model and the VARW,MSEW combination model indicates that the weight of the combination model can also choose GFW.(4)The off season daily tourist flow forecasting model of scenic spots is established.To deal with the problem that the non-linear and fluctuation characteristic of off season’s daily tourist flow are heavier than peak season,the VARW and MSEW combination model which are fixed weight combination model are difficult to grasp the fluctuation characteristics,the timing dynamic weight combination model is proposed.The model add time factor to VARW and MSEW,namely VARW-T and MSEW-T respectively.Based on the daily tourist flow of off season and relevant data in scenic spot from 2011 to 2015,three single models SVR,BP,RF and several fixed weight combination models are established.The dissertation also established several timing dynamic weight combination models for different N value of VARW-T and MSEW-T.It is found that the MAPE of the model is the smallest when the N value is about 14.The final empirical results show that the timing dynamic weight combination model is better to grasp the fluctuation characteristics of off season’s daily tourist flow than the fixed weight combination model.Therefore,the prediction effect is better and the accuracy is higher.The study of this dissertation benefits the daily management and services of the scenic spot,and also has instructional significance for the short-term tourist flow forecasting of mountain scenic spot.