摘要
为解决股指时序预测趋势难问题,提出了引入多元宇宙优化算法(Multi-Verse Optimization,MVO)的双向长短期记忆网络(BiLSTM)与卷积神经网络(CNN)相结合的股票预测混合模型(MVO-CNN-BiLSTM)。该模型的意义在于多元宇宙优化算法具有全局搜索能力和收敛速度快的特点,适用于优化问题;卷积神经网络(CNN)能够在前几层有效地提取数据中的特征,而BiLSTM则可以在后续层中建模这些特征之间的时序关系。最后通过对沪深300指数2002—2024年共5378组数据进行仿真结果表明,确定最佳时间步长后,该方法预测效果明显优于传统CNN-LSTM网络模型以及CNN-BiLSTM网络模型,能够有效降低预测误差。
In order to solve the difficult problem of predicting the trend of stock index time series,a hybrid model(MVOCNN-BiLSTM)for stock prediction is proposed,which introduces MVO,and combines BiLSTM and CNN.The significance of the model is that the MVO has the characteristics of global search ability and fast convergence speed,which is suitable for optimization problems.The CNN can effectively extract features from the data in the first few layers,and BiLSTM can model the time-series relationship between these features in subsequent layers.Finally,the simulation results of 5378 sets of data from 2002 to 2024 of CSI 300 index show that the prediction effect of this method is obviously better than that of traditional CNNLSTM network model and CNN-BiLSTM network model after determining the optimal time step,which can effectively reduce the prediction error.
作者
李梦雨
LI Mengyu(Hubei University of Technology,Wuhan 430068,China)
出处
《现代信息科技》
2024年第19期134-140,共7页
Modern Information Technology
关键词
双向长短期记忆网络
多元宇宙优化算法
卷积神经网络
Bidirectional Long Short-Term Memory network
Multi-Verse Optimization
Convolutional Neural Network
