云南民族大学学报（自然科学版）2025，Vol.34Issue(5)：572-581,10.DOI:10.3969/j.issn.1672-8513.2025.05.010
基于卷积与注意力增强的股票价格预测方法
Stock price prediction method based on convolution and attention enhancement
摘要
Abstract
The stock market is influenced by factors such as macroeconomic conditions,policy changes,and investor behavior,exhibiting high nonlinearity and dynamic complexity,making it difficult for traditional prediction models to effectively respond.In recent years,deep learning methods represented by Long Short-Term Memory(LSTM)networks have made notable progress in time series prediction.Nevertheless,they still face limitations in capturing complex feature relationships and spatial dynamics.To address these issues,this paper proposes a Temporal-Spatial Channel Long Short-Term Memory(TSC-LSTM)model that integrates Convolutional Neural Networks(CNN)with spatial and channel attention mechanisms.By incorporating convolutional feature extraction,a residual channel attention module(RCAM),and a multi-scale spatial attention module(MSAM),the proposed model enhances the representation of both local and global features in stock price data.Experimental results on datasets from Ping An Bank,Kweichow Moutai,and the Shanghai Composite Index demonstrate that the TSC-LSTM model achieves lower errors in MAE,MSE,and RMSE metrics,indicating superior prediction accuracy and generalization capability.关键词
股价预测/深度学习/LSTM/卷积神经网络/注意力机制Key words
stock price prediction/deep learning/LSTM/convolutional neural network/attention mechanism分类
信息技术与安全科学引用本文复制引用
罗云芳,张广莹..基于卷积与注意力增强的股票价格预测方法[J].云南民族大学学报（自然科学版）,2025,34(5):572-581,10.基金项目
国家社会科学基金(23BJL063). （23BJL063）
