基于ML-DMA的黄金期货价格预测研究
Gold Futures Price Forecasting Based on ML-DMA
范彩云 1童君逸 1程俊彦 1周勇2
作者信息
- 1. 上海对外经贸大学统计与信息学院,上海 201620
- 2. 统计与数据科学前沿理论及应用教育部重点实验室,上海 200062;华东师范大学经济与管理学部统计学院和统计交叉科学研究院,上海 200062
- 折叠
摘要
在黄金期货价格预测问题的研究中,价格具有时变性、非线性、高噪声和影响因子复杂等因素,决定了其被准确预测的难度.传统方法对黄金期货价格的预测主要借助于静态模型,导致预测精度不高或分析不足.为了能动态而准确的预测黄金期货价格,本文从技术行情指标、行业方面的影响因素及宏观经济环境指标三个维度选取39个变量,以机器学习(machine learning;ML)方法构建基本融合素材,利用动态模型平均(dynamic model averaging,DMA)方法代替传统模型融合技巧,得到黄金期货价格预测模型.实证结果表明,采用机器学习 动态模型平均策略能够明显提高黄金期货价格的预测精度.
Abstract
In the research of gold futures price forecasting,the price has time-varying,nonlinear,high noise and complex influencing factors,which determines the difficulty of accurate prediction.Traditional methods mainly rely on static models to predict gold futures prices,resulting in low prediction accuracy or insufficient analysis.In order to dynamically and accurately predict the price of gold futures,this paper selects 39 variables from the three dimensions of technical market indicators,industry influencing factors and macroeconomic environment indicators,constructs the basic aggregation material with machine learning(ML)method,and uses dynamic model averaging(DMA)method to replace the traditional model aggregation skills,then derive the new gold futures price prediction model.The empirical results show that the machine learning dynamic model average strategy can significantly improve the prediction accuracy of gold futures price.
关键词
时间序列预测/融合模型/动态模型平均/时变性/黄金期货价格Key words
time series prediction/hybrid model/dynamic model average/time variability/gold futures price引用本文复制引用
基金项目
国家自然科学基金面上项目(12271343)
国家自然科学基金重点项目(71931004)
培育基金(92046005)
出版年
2024
NETL
NSTL
万方数据