高师理科学刊2025，Vol.45Issue(11)：29-32,39,5.DOI:10.3969/j.issn.1007-9831.2025.11.006
基于多头注意力机制的股票趋势预测
Stock trend prediction using multi-head self-attention mechanism
代雨鑫 1马昱晗 1孙德山1
作者信息
- 1. 辽宁师范大学 数学学院,辽宁 大连 116029
- 折叠
摘要
Abstract
With the increasing complexity of financial markets,stock trend prediction has become a popular research topic.This paper integrates long short-term memory(LSTM)networks and gated recurrent units(GRU)with a multi-head self-attention(MSA)mechanism to construct two hybrid models,MSA-GRU and MSA-LSTM,aiming to enhance the accuracy of stock trend forecasting.An empirical study was conducted using data from 10 selected stocks spanning the period from 2015 to 2024.An 18-dimensional feature system was established,and the models were evaluated using accuracy,F1 score,and AUC value.The experimental results show that the MSA-GRU model achieves the best performance,with an accuracy of 84.64%and an AUC of 0.895,significantly outperforming the traditional LSTM,GRU,and MSA-LSTM models.关键词
多头注意力机制/门控循环单元/长短期记忆网络/股票趋势预测Key words
multi-head self-attention mechanism/gated recurrent units/long short-term memory networks/stock trend prediction分类
信息技术与安全科学引用本文复制引用
代雨鑫,马昱晗,孙德山..基于多头注意力机制的股票趋势预测[J].高师理科学刊,2025,45(11):29-32,39,5.
