基于STAG-Net模型的多元白酒股票价格预测  

Multivariate Baijiu stock price prediction based on STAG-Net model

作　　者：黄丹 Huang Dan(Faculty of Business Administration,Moutai University,Renhuai,Guizhou 564507,China)

机构地区：[1]茅台学院工商管理学院,贵州仁怀564507

出　　处：《计算机时代》2025年第4期59-63,共5页Computer Era

基　　金：茅台学院高层次人才科研启动经费项目(mygccrc[2024]021);贵州省高校人文社会科学研究项目(2024RW146)。

摘　　要：白酒股票价格的准确预测可辅助投资决策,提升企业风险控制与资产配置效率。传统方法大多侧重于单只股票的预测,忽视了股票间的相互影响。为此,本文结合时空模块和门控策略,提出一种新的时空预测模型STAG-Net。该模型基于GRU和空间注意力机制构建时空模块来捕捉股票之间的时空依赖性,并通过一种门控策略纳入历史均值以增强时序建模能力。实验结果表明,STAG-Net模型在多元白酒股票价格预测任务中显著优于传统的XGBoost、LSTM、GRU和Attention方法,可为金融市场中白酒板块的投资决策提供可靠的技术支持。The accurate prediction of Baijiu stock price can assist in investment decision-making and improve the efficiency of enterprise risk control and asset allocation.Traditional methods mostly focus on predicting individual stocks,ignoring the mutual influence between stocks.Therefore,this article combines spatiotemporal modules and gating strategies to propose a new spatiotemporal prediction model STAG-Net.This model is based on GRU and spatial attention mechanism to construct spatiotemporal modules to capture spatiotemporal dependencies between stocks,and incorporates historical means through a gating strategy to enhance temporal modeling ability.The experimental results show that the STAG-Net model is significantly superior to the traditional XGBoost,LSTM,GRU and Attention methods in the task of forecasting the stock price of diversified Baijiu,and can provide reliable technical support for the investment decision of Baijiu sector in the financial market.

关 键 词：GRU 空间注意力机制 时空依赖 白酒股票 价格预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] F832[自动化与计算机技术—控制科学与工程]