基于网络搜索数据的地区RPI增速预测研究——模拟河北省日度网络搜索指数对地区RPI影响分析

Research on the prediction of regional RPI growth based on web search data——the simulation of the impact of daily network search index on RPI in Hebei province

大数据时代可有效利用高频网络搜索数据预测传统宏观经济走势。基于网络搜索数据对商品零售价格指数(RPI)变动影响的分析,利用相关关键词的网络搜索数据,构建河北省地区日度网络搜索指数,并基于混频数据模型MIDAS-AR(1)对地区RPI增速进行预测。研究结果表明:基于日度网络搜索指数的混频数据模型适用于地区RPI增速的预测。在静态预测方面,采用混频数据模型的预测精度要优于同频模型。在动态预测方面,采用调整模型h值的混频数据模型的预测效果优于AR(1)模型的动态预测结果,说明构建的日度网络搜索指数对地区RPI变动具有一定的预测能力。基于此,在对宏观经济进行监测预测时,应有效利用网络海量即时数据;混频数据模型可应用于区域RPI增速预测。

High-frequency network search data can be effectively used to predict traditional macroeconomic variables in the era of big data.Based on the influence of network search data on RPI fluctuation,the daily network search index of Hebei Province is constructed by making use of the network search data of relevant keywords.And forecasting the growth of regional RPI using the MIDAS-AR(1) model.The results show that it is appropriate to predict the regional RPI using MIDAS-AR(1) model based on daily network search index.In the aspect of static prediction,the performance of models based on mixed-frequency data is better than that based on the same frequency data.In the aspect of dynamic prediction,the performance of MIDAS-AR(1) with adjusting the value of h is better than that of AR(1).The daily network search index can predict the fluctuation of regional RPI.Therefore,we should make use of the massive real-time data on the network effectively when monitoring and forecasting the macroeconomy;The mixed-frequency data model can be applied to regional RPI growth prediction.