Research On Forecasting Model Of Visitors Based On Web Search Data

In recent years,our tourism industry has developed rapidly,and travel has become the norm of life.However,tourists generally reflect that the tourism experience is gradually deteriorating.The famous scenic spots are crowded and chaotic during holidays,and travel difficulties are getting worse.The reason for this is mainly because during the peak period of tourism,the attraction capacity of scenic spots does not match with the influx of tourists,which makes the managers unprepared,causing confusion in the scenic area,destroying the scenic environment,influencing the city tourism image,and even causing security accidents.Therefore,if we can scientifically and accurately predict the amount of tourists,and use this as the basis for the preparation of emergency plans,adjust the operation and management methods of scenic spots,and then improve tourism safety and service quality,in order to promote the sustainable and healthy development of tourism.At the same time,in the age of network information,it has become one of the habits of people to go through the search for information on the Internet and to make good preparations in advance.Therefore,in the research on the prediction of the recent development of related things,web search data has gradually received the attention of researchers.On the basis of summarizing and analyzing predecessors'researches,this paper uses a more granular monthly data,taking the amount of visitors received in Beijing as an example,to explore suitable forecasting methods and innovative optimization forecasting models to achieve more accurate and timely forecasting results.The main research contents of this article include the following aspects:(1)From the point of the process of tourism behaviors,a conceptual framework is established to connect the web search data with the tourist quantity,and then based on text mining and the six elements of tourism,core keywords are obtained,and they are used as explanatory variable to establish a preliminary prediction model.Based on this,the relevance degree between the web search data and the amount of tourists is verified,and the prediction ability of the core keywords is analyzed.(2)In order to ensure the integrity of the information and improve the scientific and accuracy of prediction,the core keywords are extensively expanded to form the initial keyword thesaurus,and then the correlation keywords and the time difference analysis are used to filter out 21 high relevance keywords,at the same time,the time-delay order is determined.Ten best model variables were selected from the Adaptive-Lasso method.For comparison,principal component analysis was also used to obtain three structural variables.At this point,it is ready for the model establishment.(3)Using ELM neural network,support vector regression,and random forest algorithm,based on the 10 best variables and 3 structural variables,a number of tourist quantity prediction models were established,and the prediction ability of different models was compared and analyzed.The results show that using the optimal variables selected by the Adaptive-Lasso method to establish a model,the model's various evaluation indicators are all better.In the three machine learning prediction models,the support vector regression model is optimal,the random forest model is the most stable,and the ELM neural network model is relatively poor,in general,they all achieved better prediction results.In order to further improve the stability and generalization ability of the model,this paper introduces the ideas of the combined forecasting method to further study.The results show that the variable weight combination models can significantly improve the prediction accuracy.