摘要：Due to the impact of external objective factors, the weekly number of tourist distribution is imbalance and shows strong nonlinear characteristics, especially more obvious in the off-season. Aiming at this problem, this paper built two different forecasting models by SVR and BP neural network in order to guess about the weekly tourist numbers and then, put forward the Goodness of Fit Weight (GFW) and used it to make three SVR-BP combination models (LCM, GEOM-WTD, and HARM). In order to support the results empirically, Data of the Huangshan Scenic Area has been used and the dynamic virtual variable was adopted to deal with the objective factors that influence the weekly visitor numbers. Finally, we compare the results with the two combination models that based on the VARW and MSEW, it shows that the GFW combination model is helpful for the combination forecasting and can be used as one of selections just like other weights, and the use of dynamic virtual variables can further improve the forecasting precision.
关键词：Tourist off-season; Weekly tourists forecasting; Goodness of Fit Weight; ; Combination model
