摘要
短期微观旅游需求具有强非线性特征,单一的模型很难做出准确预测。针对此问题,本文分析了著名风景区黄山2010年旅游旺季(4-10月)相关日数据的特征,在此基础上建立SVR-ARMA组合模型,用SVR模型先对原始非线性数据预测,再对SVR模型预测所产生的线性残差用ARMA模型预测,将两部分预测值几何相加得最终的预测值。最后分别与单一的SVR和ARMA模型对比,结果表明该组合模型有更高更稳健的预测精度,很适合短期微观旅游需求。
Over the past several decades,tourism has been a key reason that the global economy has experienced such rapid development. There is a growing need for forecasting tourism demand. The current research has mainly focused on the method of model contrast and macro data rather than micro data. During high tourist season,forecasting tourism demand at the short-term,micro level is particularly important for the tourism management department. Based on the number of daily sightseers at the Mount Huangshan Scenic Spot during high tourist season（ from April to October in 2010）,this paper focuses on micro level and constructs the forecasting combination model SVR-ARMA which contains both nonlinear and linear features.In the introduction section,this paper discusses and summarizes the model and methodology used in forecasting tourism demand in the past several decades. This paper explains the source of the data used in this study. It also analyzes data features and factors that influence the number of sightseers. For the second part,the theory of SVR and ARMA is introduced,and the combination model SVRARMA is built according to data features. The SVR with a highly nonlinear processing capability is used to model and forecast because of the nonlinear feature. The residuals generated by the SVR utilizes the ARMA. The sum of the two sections forms the final predicted value. The third part is mainly about the model＇s input variables. The lags input is determined by the correlation analysis of the collected data. The input of the exogenous variables utilizes the dummy variables by analyzing data features. In the fourth part of the paper,the test of the model and the analysis of the result are mentioned. The test result indicates that utilizing SVR-ARMA is more effective than using SVR and ARMA separately. The conclusion and future research directions are made in the last part of the paper.Different from other prediction literatures,this paper constructs a predictive model based on the analysis of data features instead of te
出处
《管理工程学报》
CSSCI
北大核心
2015年第1期122-127,共6页
Journal of Industrial Engineering and Engineering Management
基金
国家自然科学基金资助项目(71271072
71201045)
高等学校博士点基金资助项目(20110111110006)
安徽高校省级自然科学研究资助项目(KJ2012B097)
智慧黄山风景区人流量智能分析预测系统资助项目(10120106011)
