Research On Tourism Demand Forecasting Model Based On Multi-source Data

In the Internet plus era,accurate prediction of tourism demand has become the biggest challenge for tourism specialty people and research scholars.And because the tourist number has complex nonlinearity feature,all the time it is a key issue in study of tourism demand forecasting.Abundant search data and interactive data become favorable source for predicting tourist number.Otherwise,prediction of tourist number is influenced significantly by weather,leading to the problem that modeling for predicting tourist number is difficult.This article regards Hainan Province of large scale tourism destination and Xi’an Museum of small scale tourism scenic spot as object of study,takes prediction of tourist number as key point of research objective and explains secondarily prediction of tourism income which is a supplement for the full text.The main content of this research is as follows:(1)By analyzing the influence factors this paper proposed a conceptual prediction framework of three layers.Based on the development of Hainan Province of large scale tourism destination and Xi’an Museum of small scale tourism scenic spot,this article analyzes influence factor of tourist flow from time’ dimension and spatial dimension;systematically studies evolution law of tourist flow,structural feature,space distribution and so on from the angles of different time scale and tourist market.To further predict tourist flow,taking data-driven method as theory foundation,this article uses basic data,search data and interactive data to build a model for predicting tourist flow based on multy-source data.This article utilizes grey relational analysis method and correspondence analysis method to conduct complementary analysis of influence factors;realizes a conceptual prediction framework of three layers from the levels of data,analysis and model method;finally proposes tourist flow change modeling scheme.(2)Aiming at Hainan Province tourism reception situation and the data source of general income,due to the feature of nonlinear fluctuation,considering prediction algorithms of Grey Model and Markov Model,this article proposes an optimal input subset based on dynamics and DFS-GKM,which means Dynamic Fuzzy Optimal Input Subset Grey Markov Model.Aiming at Hainan Province tourism reception situation and the data source of general income,due to the feature of nonlinear fluctuation,considering prediction algorithms of Grey Model and Markov Model,this article proposes a DFS-GKM method based on optimal input subset and fuzzy theory,which means Dynamic Fuzzy Optimal Input Subset Grey Markov Model.To begin with,the article utilizes the Optimal Input Subset Method and relies on traditional GM model prediction result average absolute error percentage to optimize input subset and confirm the most optimal input subset sequence.Then the article uses fuzzy set to calculate degree-of-membership vector which is the weight of Markov transition matrix vector,modified predicted value.Finally,to predict based on time lapse,this article establishes equivalent dimensions dynamic forecasting model.Based on the experiment data from Hainan Province of large scale tourism destination,this article states it can efficiently improve prediction accuracy of oscillatory sequence conducted by Grey Model.Additionally,the SVM regression prediction method is used to contrast mean square error and relative coefficient of general income forecasting about different data sets;to get the conclusion that for the mixed data set,mean square error is lowest,relative coefficient is highest,prediction accuracy is highest.Then the article uses fuzzy set to calculate degree-of-membership vector which is the weight of Markov transition matrix vector,modified predicted value.Finally,to predict based on time lapse,this article establishes equivalent dimensions dynamic forecasting model.Based on the experiment data from Hainan Province of large scale tourism destination,this article states it can efficiently improve prediction accuracy of oscillatory sequence conducted by Grey Model.(3)Aiming at the data source of the number of ticket holder entering Xi’an museum with obvious randomness and volatility feature,this article proposes a kind of combination optimization model based on fuzzy time series model and grey model.This article uses EM to give proper weight to two kinds of single forecasting model;applies entropy method into confirmation of weighted average of combination model;for Markov Model,based on its no aftereffect feature which is suitable for data sequence of large volatility,prediction value of combination optimization model can be modified by Markov Model.The experiment result based on tourist flow of Xi’an Museum of small scale tourism scenic spot testifies combination optimization model is obviously better than single model,which improves accuracy efficiently.This example shows that combination optimization model reduces damage of information during fusing and enhances fusion accuracy.