基于Attention机制的CNN-LSTM概率预测模型的股指预测  

Stock Index Prediction Based on CNN-LSTM Probability Prediction Model with Attention Mechanism

作　　者：高欣 GAO Xin(School of Big Data and Statistics,Anhui University,Hefei 230601,China)

机构地区：[1]安徽大学大数据与统计学院,安徽合肥230601

出　　处：《现代信息科技》2024年第12期155-159,163,共6页Modern Information Technology

摘　　要：鉴于证券市场波动大预测难度高,文章基于encoder-decoder结构将Attention机制融入CNN-LSTM模型,利用Attention机制来捕捉不同时间点之间的数据依赖模式,提取长序列信息,并且在此基础上给出概率密度函数进行抽样预测,最终得出股票价格的点预测和区间预测。实验结果表明,融入Attention机制的CNN-LSTM概率预测模型从综合性能来看优于其他基准模型,能够对上证指数收盘价进行较高精度的多步预测。Given the high volatility of the securities market and the high difficulty of predicting it,this paper integrates the Attention Mechanism into the CNN-LSTM model based on the encoder-decoder structure.The Attention Mechanism is used to capture data dependency patterns between different time points,long series information is extracted,and based on this,a probability density function is provided for sampling prediction,point prediction and interval prediction of stock prices are obtained ultimately.The experimental results show that the CNN-LSTM probability prediction model incorporating the Attention Mechanism outperforms other benchmark models in terms of comprehensive performance,and can make high-precision multistep predictions of the closing price of the Shanghai Composite Index.

关 键 词：Attention机制 概率密度函数 上证指数 

分 类 号：TP39[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]