统计研究 ›› 2020, Vol. 37 ›› Issue (7): 104-115.doi: 10.19343/j.cnki.11-1302/c.2020.07.009

• • 上一篇    下一篇

基于机器学习LSTM&US模型的消费者信心指数预测研究

唐晓彬 董曼茹 张瑞   

  • 出版日期:2020-07-25 发布日期:2020-07-15

Research on the Prediction of Consumer Confidence Index Based on Machine Learning LSTM&US Model

Tang Xiaobin Dong Manru Zhang Rui   

  • Online:2020-07-25 Published:2020-07-15

摘要: 消费者信心指数等宏观经济指标具有时间上的滞后效应和动态变化的多维性,不易精确预测。本文基于机器学习长短时间记忆(Long Short-Term Memory,LSTM)神经网络模型,结合大数据技术挖掘消费者信心指数相关网络搜索数据(User Search,US),进而构建一种LSTM&US预测模型,并将其应用于对我国消费者信心指数的长期、中期与短期的预测研究,同时引入多个基准预测模型进行了对比分析。结果发现:引入网络搜索数据能够提高LSTM神经网络模型的预测性能与预测精度;LSTM&US预测模型具有较好的泛化能力,对不同期限的预测效果均较稳定,其预测性能与预测精度均优于其他六种基准预测模型(LSTM、SVR&US、RFR&US、BP&US、XGB&US和LGB&US);预测结果显示本文提出的LSTM&US预测模型具有一定的实用价值,该预测方法为消费者信心指数的预测与预判提供了一种新的研究思路,丰富了机器学习方法在宏观经济指标预测领域中的理论研究。

关键词: 机器学习, 网络搜索数据, LSTM&US预测模型, 消费者信心指数

Abstract: Macroeconomic indicators such as the consumer confidence index have the time-lagging effect and the multidimensional dynamic changes,making it difficult to predict accurately.Based on machine learning Long Short-Term Memory (LSTM) neural network model,combined with big data technology to mine User Search (US) data related to consumer confidence index,this paper builds an LSTM&US prediction model.It is then applied to the long-term,medium-term and short-term forecasting research of China′s consumer confidencen index,and multiple benchmark prediction models are introduced for comparative analysis.The results show that the introduction of User Search data can improve the prediction performance and accuracy of the LSTM neural network model;the LSTM&US prediction model has a good generalization ability,and the prediction effect is stable for different periods. Its prediction performance and prediction accuracy are better than the other six benchmark prediction models(LSTM,SVR&US,RFR&US,BP&US,XGB&US,and LGB&US).The prediction results show that the LSTM&US prediction model proposed in this paper has a certain practical value,and the prediction method provides a new idea of forecast and anticipation for consumer confidence index,enriching the theoretical research of machine learning methods in the field of macroeconomic index prediction.

Key words: Machine Learning, User Search Data, LSTM&US Prediction Model, Consumer Confidence Index