基于互联网大数据背景下的CPI预测研究

Research Based on The Background of Big Data in The CPI Prediction

本文基于网络搜索数据,以宏观层面、微观层面和金融层面为研究视角,选取与CPI(consumerpriceindex,居民消费价格指数)相关的网络搜索词,获取其百度指数,运用PC A(principal component analysis,主成分分析)降维方法合成宏观类综合指数、微观类综合指数和金融类综合指数。以该三类综合指数为基础,建立CPI预测模型并拟合出CPI的预测值。结果显示,基于网络搜索数据拟合出的CPI预测值与真实的CPI的走势相吻合,样本内和样本外预测结果的平均绝对误差均较小。本文预测CPI的方法能克服传统CPI统计方法的滞后性,可以在一定程度上提前预测出CPI的走势及其拐点,可为未来宏观经济形势预测提供参考。

In this paper,from the perspective of web search,we select web-search key words related to CPI in macroscopic,microscopic and financial aspects,acquire its Baidu Index,use PCA to implement dimension reduction and compound macro comprehensive index,the micro comprehensive index and financial comprehensive index.Then based on these three indexes,we establish CPI forecasting model and obtain its fitted value.The result shows that the fitted value on account of web search corresponds to the real CPI very well.Mean absolute error of both forecasting in sample and forecasting out of sample are pretty slight.In addition,the forecasting method in this paper can overcome the hysteresis of traditional forecasting methods.