| As the top five happiness industries,the high-quality development of the tourism industry plays an important role to achieve people’s aspirations and goals for a better life.The "14th Five Year Plan for the Development of Tourism Industry" issued by the State Council points out that under the normalization of epidemic prevention and control,the main attention should be on deepening the supply side structural reform of the tourism industry and emphasize ’emphasizing demand side management’.Therefore,adopting the comprehensive research methods to improve the accuracy of tourism demand forecasting and actively promoting the practical application of tourism demand forecasting research has received widely attention from scholars.However,due to some inherent characteristics of the tourism industry,it is difficult to forecast the tourism demand accurately.On the one hand,the tourism industry is susceptible to various internal and external factors,leading to severe fluctuations of tourism demand and resulting in non-stationary and nonlinear characteristics of the tourism demand time series.On the other hand,the diversification of tourism demand forecast objects requires researchers to design corresponding forecast models,and this process is often time-consuming and labor-intensive.Therefore,to better address this complex and diverse tourism demand forecast problem,this paper proposes a sub-series adaptive forecasting strategy based on decomposition-ensemble framework.This strategy can reduce the complexity of the time series through decomposition methods and solve the diversification problem in tourism demand forecasting by improving the forecast strategy of the decomposition-ensemble model.Specifically,this strategy can simultaneously consider the different data characteristics of the sub-series and the relationship between the sub-series forecast results and adaptively match the corresponding forecast model for each sub-series and determine the model structure.Therefore,it can effectively overcome the problems of forecasting strategies in existing research,such as ignoring different data features of sub-series and insufficient generalization ability and thereby improving the overall forecasting performance of the decomposition-ensemble model.Due to the adaptive and high-dimensional characteristics of the proposed model,this study combines the coevolutionary framework for high-dimensional optimization problems with the heuristic algorithm(differential evolution algorithm),and proposes the cooperative coevolution differential evolution algorithm(CC-Co DE)to optimize the parameters in the proposed model.In addition,in order to test the validity and stability of the proposed model,this study selected 16 countries and scenic spots as experimental cases which including seven famous scenic spots in Shanghai and nine major tourist source countries of Hong Kong,based on different time scales(monthly and hourly)and spatial scale(countries and scenic spots),and determined nine benchmark models from three aspects.Then,we use the proposed model and each benchmark model to forecast the experimental cases.Through the analysis of the results of two experiments,the experimental results show that the proposed adaptive sub-series forecasting strategy based on decomposition-ensemble has significant advantages. |
