基于客源地的聚类-ARIMA模型的短期旅游需求预测--以天津欢乐谷主题公园为例  

Short-term Forecasting Tourism Demand Based on Origin's Hierarchical Clustering ARMA Model: A Case Study of Tianjin Happy Valley Theme Park

作　　者：段莉琼[1,2] 宫辉力[1] 刘少俊[3] 刘泽华[4] 李勇永 葛军莲[3] 

机构地区：[1]首都师范大学资源环境与旅游学院,北京100048 [2]国家旅游局信息中心,北京100740 [3]南京师范大学地理科学学院,南京210023 [4]南京大学地理与海洋科学学院,南京210093

出　　处：《地域研究与开发》2017年第3期108-112,141,共6页Areal Research and Development

基　　金：国家自然科学基金项目(41671145,41301144);国家旅游局青年专家培养计划项目(TYETP201312)

摘　　要：大多数旅游需求预测研究是基于目的地游客总数或消费总量开展的,尚未按不同的旅游目的或客源地细分进行预测。以天津欢乐谷主题公园为案例地,选择2014年第40周到2015年第26周为研究时段,利用通信大数据,提出了一种面向客源地的聚类-ARIMA组合预测模型。通过对不同客源地的时序数据进行聚类,选取各类别中的代表性客源地分别构建ARIMA预测模型。结果表明:对欢乐谷主题公园各客源地分别建模与聚类后通过6个代表客源地建模得到的结果一致;后者可以降低80%的预测成本。该方法具有较高的预测精度和较低的计算成本,适合面向客源地的短期旅游需求预测,可为旅游目的地提供更具针对性的旅游需求管理、分析与决策支撑。Most of the tourism demand forecast studies are based on the total number of tourists or the total consumption of the destination, and has not yet been forecasted according to different destination or tourist origin. This paper proposes a time-series clustering-ARIMA for the tourist origin based on the large data of communication by selecting the Tianjin Happy Valley Theme Park as a case and taking the 40th week of 2014 to the 26th week of 2015 as a studied time period. The time series data of the different tourist origin are classified by time series cluste- ring method.The typical tourist origin of different type are selected to build the respective ARIMA prediction model. The results show that the modeling of the typical tourist origin after clustering are consistent with the modeling of the different tourist origin. The cost of forecasting can be reduced by 80% by the modeling of the typical tourist origin. This method has higher prediction accuracy and lower computational cost. It is suitable for short-term tourism de- mand forecasting and can provide more targeted tourism demand management, analysis and decision support for tourism destination.

关 键 词：短期旅游需求预测 客源地 时间序列聚类 ARIMA模型 天津欢乐谷主题公园 

分 类 号：F590[经济管理—旅游管理]