大数据方法与宏观经济监测预测应用
Big Data Methods and Their Applications in Macroeconomic Monitoring and Forecasting
郑挺国 1范馨月2
作者信息
- 1. 厦门大学宏观经济研究中心/邹至庄经济研究院/经济学院统计与数据科学系 厦门市思明区厦门大学经济学院 361005
- 2. 上海财经大学经济学院 200433
- 折叠
摘要
大数据方法在宏观经济监测与预测中的创新应用意义重大.在当前全球经济不确定性增强的背景下,传统宏观经济监测预测手段面临时效性和准确性不足的挑战,而大数据技术为宏观经济分析与预测提供了前所未有的新机遇.本文首先阐述了大数据挖掘与处理的关键步骤和技术方法,特别是针对新闻媒体文本数据这一重要非结构化数据源的处理,通过主题模型及大语言模型等方法提取有效信息以构建经济监测预测指标;其次介绍了大数据驱动的宏观经济监测方法,重点讨论了混频动态因子模型及其扩展应用,这些模型能够有效处理多频率、多来源的复杂数据,实现宏观经济的实时跟踪与监测;最后探讨了大数据在提升宏观经济关键指标预测精度方面的潜力,以及大数据技术对宏观经济风险进行监测与识别的可能性,特别是通过大语言模型等人工智能技术捕捉文本数据中的语义信息和情感倾向,进行更为精准的预测.
Abstract
We explore the innovative application of big data methods in macroeconomic monitoring and forecasting,emphasizing that in the current context of increasing global economic uncertainty,traditional macroeconomic monitoring and forecasting methods face challenges in terms of timeliness and accuracy,while big data injects new vitality into macroeconomic analysis.For this purpose,we first elaborate on the key steps and technical methods of big data mining and processing,including data cleaning,representation,and structured processing,especially for the processing of important unstructured data sources such as news media textual data.Through methods such as the topic method or Large Language Model,we can extract effective information to construct economic monitoring and forecasting indicators.Secondly,we introduce the macroeconomic monitoring methods driven by big data,with a focus on discussing the mixed-frequency dynamic factor model and its extended applications.These models can effectively handle multi-frequency and multi-source data,achieving real-time tracking and monitoring of macroeconomics.Finally,we explore the potential of big data in improving the accuracy of macroeconomic predicting,particularly by using artificial intelligence technologies such as large language models to capture semantic information and emotional tendencies in textual data for more precise predictions.
关键词
大数据/文本分析/混频数据/混频动态因子模型/MIDAS模型/大语言模型/宏观经济预测Key words
Big Data/Text Analysis/Multi-frequency Data/Mixed-Frequency Dynamic Factor Model/MIDAS Model/Large Language Model(LLM)/Macroeconomic Forecasting引用本文复制引用
出版年
2024