高频数据与大数据方法在通胀预测中的应用  

Application of High Frequency Data and Big Data in Inflation Prediction

作　　者：熊伟琪 张鹏[1] 

机构地区：[1]中信银行总行金融市场部

出　　处：《金融市场研究》2022年第12期94-103,共10页Financial Market Research

摘　　要：在近几年国际国内超预期事件频发的市场环境中,本文从通胀预测的现实需求出发,在建立稳定的单变量模型的基础上,加入高频指标,增加了模型预测的准确性和及时性。同时,在高频指标的选择部分,我们应用大数据降维技术,采用主成分分析方法选取更有价值的预测指标。最后,简要讨论了在疫情扰动、基期轮换、春节效应等特殊情形下通胀模型的预测表现。This paper seeks to address the demand for inflation forecasting. On the basis of establishing a stable univariate model, it adds high-frequency indicators, which increases the accuracy and timeliness of the model’s forecasts. At the same time, it selects the dimension reduction technology of big data as high-frequency indicators and adopts the principal component analysis method to select more valuable prediction indicators. Lastly, it discusses the prediction performance of inflation models under special circumstances such as an epidemic disturbance, base period rotation and the Spring Festival effect.

关 键 词：通胀预测 大数据方法 特征缩减技术 主成分分析 

分 类 号：F822.5[经济管理—财政学] TP311.13[自动化与计算机技术—计算机软件与理论]