| 摘 要: 文章以2012—2022年上证50、沪深300、中证500三支股指交易数据为研究对象,利用深度学习预测股指,分别构建了卷积神经网络(CNN)模型、长短期记忆神经网络(LSTM)模型和CNN-LSTM组合神经网络模型,使用均方根误差(RMSE)、平均绝对误差(MAE)、平均百分比误差绝对值(MAPE)和决定系数(R2)4个指标对模型进行评价。结果表明,组合模型兼有CNN模型的高维特征挖掘优势和LSTM 模型的时序特征提取优势,在3支股指的预测中,组合模型的预测精度均高于CNN模型和LSTM 模型单一模型,其RMSE指标分别提升了28.60百分点、52.56百分点和25.28百分点,证实了CNN-LSTM组合神经网络模型的有效性和准确性。 |
| 关键词: 深度学习;股指预测;CNN模型;LSTM模型;CNN-LSTM组合神经网络模型 |
|
中图分类号: TP183
文献标识码: A
|
|
| Stock Price Index Prediction Based on Deep Learning |
|
GAO Yuan, HUANG Wei
|
(School of Economics, Nanjing University of Posts and Telecommunications, Nanjing 210023, China)
2249139525@qq.com; wei.huang@njupt.edu.cn
|
| Abstract: Taking trading data of SSE50 (Shanghai Stock Exchange 50), CSI300 (China Securities Index 300), and CSI500 (China Securities Index 500) stock indexes from 2012 to 2022 as the research objects, this paper proposes to use Deep Learning to predict stock indexes. Convolutional Neural Network (CNN) model, Long Short-Term Memory Neural Network (LSTM) model, and CNN-LSTM combined neural network models are constructed to predict stock indexes. The model is evaluated using four indicators: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R2 ). The results show that the combined model has the advantages of high-dimensional feature mining of CNN model and temporal feature extraction of LSTM model. In the prediction of the three stock indexes, the prediction accuracy of the combined model is higher than that of the CNN and LSTM single models. Its RMSE indicators have increased by 28.60 percentage points, 52.56 percentage points, and 25.28 percentage points respectively, verifying the effectiveness and accuracy of the CNN-LSTM combined neural network model. |
| Keywords: Deep Learning; stock index prediction; CNN model; LSTM model; CNN-LSTM combined neural network model |
