【摘要】 网络搜索大数据为研究游客量预测提供了新的视角，而多数研究运用的传统计量经济模型难以处理网络搜索与客流时序中包含的大量非线性波动特征，导致预测精度不够理想.引入经验模态分解方法(empirical mode decomposition, EMD)将向量自回归(vector autoregression, VAR)模型改进为EMD-VAR模型. EMD方法分解夫子庙景区长三角日际网络搜索和游客量序列，得到不同频率尺度的分量，基于波动关联的视角将同一尺度的两类序列分量组合建立EMD-VAR模型进行预测.结果表明：(1)网络搜索波动周期比游客量波动周期长.(2)网络搜索与游客量波动的关联紧密度在法定节假日时期最高.(3)EMD-VAR模型比ARMA模型和VAR模型具有更高的预测精度.更多还原

【Abstract】 Big data from network search provides a new perspective for the study of tourist flow volume prediction, but the traditional econometric models used in most studies are difficult to deal with the large number of nonlinear fluctuation characteristics in the timing series of network search and tourist flow, which leads to the unsatisfactory prediction accuracy. In this paper, empirical mode decomposition(EMD)is introduced to improve the vector autoregression(VAR)model to EMD-VAR model. EMD method is used to decompose the daily network search data and tourist flow volume of The Yangtze River Delta of Nanjing Confucius Temple Scenic Area, and a series of components with different frequency scales are obtained. Then, based on the perspective of fluctuation correlation, components of both network search data and tourist flow volume in the same scale are combined to establish a VAR model for prediction. The results show that:(1)The fluctuation cycle of network search is longer than that of tourist flow volume.(2)The compactness of correlation between network search and tourists flow volume is the greatest during the statutory holiday period.(3)The prediction accuracy of the EMD-VAR model is better than that of ARMA model and VAR model, respectively.更多还原