Forecast Of Shanghai Disney's Visitor Volume Based On Baidu Index

Living condition of Chinese citizen has been raised up substantially after 40 years of revolution and opening-up.Travel industry has developed rapidly.Statistic shows that China has already become world's No.1 guest source country for outbound travel and No.4 guest reception country for inbound travel.Number of domestic travelers has reached 5 billion in 2017.Travel industry has already become a major comprehensive industry of the nation."Travel Industry Development Planning of the Thirteenth Five Years" has specified even higher requirements for the travel industry.It not only requires that growth rate of number of travelers of our citizens should be higher than 10%,but also states that travelers of our citizens should be more satisfied with the travel experience.However,famous scenic spots and cities faces higher and higher reception volume with desire of travel keep rising,especially in long holidays.Overloading guest volume not only does great damage to our ecological environment,but also leads to loss of lives.The tragic incident on Shanghai Bund in 2014 and on Huashan Shaanxi in 2012 are still in our memory.Thus,in order to avoid such tragic incidents,and at the same time the irredeemable damage to our ecological environment,it is import to study the rational and scientific method to predict the travel demand of scenic spots and cities.The paper introduces the rationale behind establishing models to predict tourist volume of scenic spots.Tourist tend to use search engine to look for information of the scenic spots and destination in order to improve travelling experience today when Internet is highly developed.This makes the foundation of establishing models to predict tourist volume based on search volume on Internet.The paper uses Baidu Index because Baidu is the most used search engine in China at the moment.The paper also introduces the basic theory of the forecasting models and model examine methods which are used in the paper.We choose daily attendance volume of Shanghai Disney resort as research object.The property of the time series is studied and we choose to use the logarithm series to build model.First of all,the original series is divided into in-example data and out-of-example data.We build ARMA,ARDL and SVR models to fit in-example series and compare their fitting result.When we build ARDL model,we need choose suitable key words and check whether the cointegration relationship and Granger Causality exists between time series of the key words and attendance volume.Then we build ARDL model based on Baidu index of the chosen key words,and import in dummy variables including weather and holiday(Saturday,Summer,3-day holiday,7-day holiday)to improve the fitting result.Then we use the regressors of the built ARDL model as input to build SVR model.At last the out-of sample rolling forecast method is used to predict for out-of-sample data.And we use three loss functions to evaluate the prediction results.We found out that the fitting result of ARDL model is evidently better than ARMA model.At the same time,the predictive ability of SVR model is evidently better than ARDL model and ARMA model.