| 利用互联网大数据预测季度GDP增速的方法研究 |
| |
| 引用本文： | 何强 董志勇. 利用互联网大数据预测季度GDP增速的方法研究[J]. 统计研究, 2020, 37(12): 91-104. DOI: 10.19343/j.cnki.11-1302/c.2020.12.007 |
| |
| 作者姓名： | 何强 董志勇 |
| |
| 摘 要： | 大数据为季度GDP走势预测创新研究带来重要突破口。本文利用百度等网站的互联网大数据,基于代表性高维数据机器学习(和深度学习)模型,对我国2011-2018年季度GDP增速深入进行预测分析。研究发现,对模型中的随机干扰因素作出一定分布的统计假设,有助于降低预测误差,任由模型通过大量数据机械地学习和完善并不总是有利于模型预测能力的提升;采用对解释变量集添加惩罚约束的方法,可以有效地处理互联网大数据维度较高的棘手问题;预测季度GDP增速的最优大数据解释变量集的稳定性较高。
|
| 关 键 词： | 互联网 大数据 季度GDP 高维 机器学习 |
Research on the Methods of Forecasting Quarterly GDP Growth Using Internet Big Data |
| |
| He Qiang Dong Zhiyong. Research on the Methods of Forecasting Quarterly GDP Growth Using Internet Big Data[J]. Statistical Research, 2020, 37(12): 91-104. DOI: 10.19343/j.cnki.11-1302/c.2020.12.007 |
| |
| Authors: | He Qiang Dong Zhiyong |
| |
| Abstract: | Big data brings an important breakthrough for innovative research on quarterly GDP prediction.Based on representative high-dimensional data machine learning (and deep learning) models, this paper uses Internet big data from Baidu and other websites to forecast China’ s quarterly GDP growth in 2011-2018 thoroughly. In this paper, we illustrate that (i) the statistical hypothesis of random interference factors in the model helps to reduce the prediction error, and allowing the model to learn and improve mechanically from a large amount of data is not always conducive to the prediction improvement; (ii) the method of adding penalty constraints to the set of explanatory variables can effectively deal with the high-dimensional Internet big data; and (iii) the optimal set of explanatory variables of forecasting quarterly GDP growth is very stable. |
| |
| Keywords: | Internet Big Data Quarterly GDP High-Dimensional Machine Learning |
|
| 点击此处可从《统计研究》浏览原始摘要信息 |
|
点击此处可从《统计研究》下载全文 |
|
