基于机器学习的多模型黄金股票价格预测  

Dynamic Trend Analysis of Gold Prices Based on Machine Learning Models

作　　者：倪文周 田嫚 焦一清 茹慧英 NI Wenzhou;TIAN man;JIAO Yiqing;RU huiying(School of Science,Hebei University of Architecture,Zhangjiakou 075000,Hebei,China)

机构地区：[1]河北建筑工程学院理学院,河北张家口075000

出　　处：《长江信息通信》2025年第2期169-172,共4页Changjiang Information & Communications

基　　金：2024年河北省科技特派员专项基金,2024—2025年大学生创新创业基金资助课题(项目编号:S202410084027)。

摘　　要：黄金价格股票作为经济分析领域的重要指标,对国民经济的盈亏起着风向标的作用。黄金价格预测模型有助于国民在市场经济下能够占据提前避险优势。文章分别选用时间序列ARIMA模型、集成学习算法中的XGBoost模型以及深度学习中的LSTM模型对黄金股票价格进行预测。通过对比三者的RMSE、MAE、MAPE三项模型指标,发现应用中的LSTM时间序列模型更优,其中,ARIMA模型三项指标为2.62、2.33、0.83%;XGBoost模型为6.161、4.943、0.83%;LSTM模型为2.194、1.789、0.73%。为了进一步提高预测精度,研究最终采用了更深层的网络架构以及双向LSTM模型。最终实现了RMSE、MAE、MAPE的进一步降低,达到1.337、1.057、0.42%,标志着模型预测能力的显著提升,可更精准地捕捉黄金价格的动态。As a pivotal indicator in economic analysis,gold prices play a barometer role in national economic fluctuations.Predictive models for gold prices enable proactive risk mitigation in market economies.This study comparatively analyzes three predictive models:the ARIMA time series model,the XGBoost ensemble learning algorithm,and the LSTM deep learning model,evaluated through RMSE,MAE,and MAPE metrics.Results demonstrate the superior performance of the LSTM model(2.194,1.789,0.73%)compared to ARIMA(2.62,2.33,0.83%)and XGBoost(6.161,4.943,0.83%).To enhance prediction accuracy,we implemented deeper network architectures and bidirectional LSTM,achieving optimized metrics of 1.337,1.057,and 0.42%respectively.This advancement significantly improves dynamic gold price pattern recognition,offering enhanced predictive capabilities for economic decision-making.The methodological progression from traditional statistical models to sophisticated neural architectures underscores the effectiveness of deep learning approaches in financial time series forecasting.

关 键 词：金价预测 ARIMA模型 XGBoost模型 双向LSTM时间序列模型 

分 类 号：O141.41[理学—数学]