Abstract: To address the problem that the characteristics of non-stationary, non-linearity, highly complex, and stochastic fluctuations of stock prices make it difficult to predict, a stock price prediction method based on the E-V-ALSTM hybrid depth model is proposed. The decomposition of the stock price data is performed using empirical modal decomposition（EMD） to obtain several intrinsic modal functions（IMFs） and a residual（Res） to reduce the non-stationarity and non-linearity of the stock price data. the complexity of these IMFs is assessed using sample entropy（SampEn）, then the IMFs with a complexity higher than a certain threshold are decomposed using variational modal decomposition（VMD） to reduce the complexity of the stock price data. A long short-term memory neural network （LSTM） model incorporating an attention mechanism is used for prediction, capturing information on key time point features and reassigning weights to address the stochastic volatility of the stock price data to improve the accuracy of the prediction method. The experimental results on datasets such as the CSI300 index and German DAX index show that the model can further improve the accuracy of stock price prediction than other comparative models.

Key words: stock price prediction, quadratic decomposition, sample entropy, attention mechanism, long short-term memory neural network（LSTM）

摘要： 针对股票价格非平稳、非线性、高复杂和随机波动等特性使其预测难度大的问题，提出一种基于E-V-ALSTM混合深度模型的股票价格预测方法。使用经验模态分解（EMD）对股票价格数据进行第一次分解，得到若干固有模态函数（IMFs）和一个残差（Res），降低了股票价格数据的非平稳性和非线性；使用样本熵（SampEn）对这些IMFs进行复杂性评估；将复杂度高于一定阈值的IMFs使用变分模态分解（VMD）进行二次分解，以降低股票价格数据的复杂性；通过加入注意力机制的长短期记忆神经网络（LSTM）模型进行预测，捕捉关键时间点特征信息，重新赋予权重，以解决股票价格数据的随机波动性，提升预测方法的精确度。对沪深300指数和德国DAX指数等数据集上的实验结果表明，该模型比其他对比模型能进一步提高股票价格预测的准确性。

关键词: 股票价格预测, 二次分解, 样本熵, 注意力机制, 长短期记忆神经网络（LSTM）