| Geographical location and traffic condition are important aspects for tourism,the sudden influx of tourists often makes people feel crowded and querulous,to which scenic managements are helpless.Therefore,it is a hot topic of public and tourism management about how to timely warn,schedule and allocate limited tourism resources.The most direct way is to forecast tourist flow,but data of current tourism statistics and inferences are mainly from the travel agency,scenic area,surrounding hotels and other departments afterwards.These data need layers of summary and submission,are usually released by month or quarter with a time lag.And the granularity of data is too large to predict the tourist flow timely and accurately.With the rapid development and maturity of Internet and online travel services,people’s access to information is being diverted from traditional channels such as TV,newspapers,and oral communication to information channels such as Internet.Massive data is constantly being generated and updated,but these data are huge and scattered,and it seems to be a needle in a haystack for information users.The search engine uses the query technology to retrieve and process information and then present it to users,its birth enables people to obtain information in a very short time and at a very low cost.Web search data from search engine records the needs and preferences of tourists before the trip,which is advanced and real-time record of tourists’ behavior and does not depend on respondents’ motivation and degree of coordination.Based on this,network search data as a high-quality information source reflecting public behavior has important auxiliary value for the study of socio-economic issues.Previous studies on the prediction of tourist flow less used network search data,or introduced network search data as a variable into a traditional time series model or an econometric model.These studies generally rely too much on historical data of tourist flow,and do not consider the short-term impact of holidays and special events,so it is difficult to make timely and accurate predictions and warnings before the surge in tourist flow.This paper takes Xi’an as an example,which is the starting point of the ancient Silk Road and popular destination for humanistic tourism,first constructs the regional tourism index based on tourist information search data.Then the paper study the relationship between index and tourist flow from both theoretical and data perspective.Furthermore,discussing the microscopic dynamic characteristics and predictive ability of index on the basis of considering the effect of holidays.The conclusion is that the regional tourism index can be used to reflect the change of passenger flow,and the Prophet model with holiday effect performs best.The full text mainly includes the following aspects:(1)This paper discusses the decision-making process of tourism consumers,constructs the theoretical framework between network search data and actual tourist flow through the consumer behavior model and the information search theory in economics,and holds that there is a positive correlation between the network search volume and the actual tourist flow.(2)With the help of hot word search volume provided by Baidu search engine,this paper sets the initial keywords and expands keywords through direct and range word search methods,screens out hot tourist words related to Xi’an,and crawls the search volume by crawling data.(3)Calculating the correlation coefficient between the normalized keyword search volume and the actual tourist flow,12 final keywords are selected and the regional tourism index is constructed by weighted summing the search volume according to the degree of correlation.(4)This paper describes the statistical characteristics of the daily regional tourism index and discusses its holiday characteristics and precursory effects by defining the "tourism concern period" indicator.(5)Innovatively using the HEGY seasonal cointegration test to verify the cointegration relationship between the regional tourism index and the actual tourist flow,the paper finds that they both have the same seasonal effect,with synchronization and consistency.To estimate the seasonal tourism index,the paper uses semi-parametric mixed model,which indicates that it has a trend consistency with the X12 seasonally adjusted tourist flow sequence,thus verifying the nowcast predictive ability of the regional tourism index.(6)Finally,this paper uses Facebook’s open source Prophet prediction model to predict daily regional tourism index series,by finding the maximum posteriori estimation with the optimization iterative algorithm,which is more flexible and accurate than traditional time series models.Furthermore,the paper introduces holiday effect into the Prophet prediction model,significantly reduces the error of fitting and prediction by 40%.At the same time,the model decomposes the regional tourism index and analyzes its trend component,week effect,annual effect and holiday effect.(7)The regional tourism index series is predicted by the classical SARIMA model and the ARIMA model after EMD decomposition.The decomposition significantly improves the prediction accuracy.(8)In order to improve the credibility of the analysis,this paper conducted a questionnaire survey on tourism behavior and retrieved 200 valid questionnaires.The questions involved demographic characteristics,tourism ’information search behaviors and tourism behavior preferences of the respondents.The conclusions are partially used for reference in keyword selection and precursor effect analysis.The shortcomings of this paper are as follows:(1)The actual situation is that some search behaviors may not lead to actual tourism activities.In the future research,the validity of network search data should be considered and filter out irrelevant information.(2)In this paper,the specific characteristics of the precursory effect of network search data are not analyzed in depth,it can be considered to distinguish the network search data between PC end and mobile to study. |
PDF Full Text Request