基于WD-CNN-LSTM模型的股票价格预测分析

Stock Price Prediction Analysis Based on WD-CNN-LSTM Model

神经网络模型一直是解决金融领域量化投资问题的重要方法。为了提高股票预测准确率,增强股票价格预测的有效性与稳定性,本研究融合卷积神经网络、长短期人工神经网络模型以及小波去噪,构建了WD-CNN-LSTM模型。该模型首先使用小波阈值去噪将股票数据进行滤波分解,提取出低频数据,然后将处理后的数据通过卷积神经网络进行特征提取,最后利用长短期人工神经网络模型对提取的特征信息进行处理,得到预测的股票价格。实验结果表明:WD-CNN-LSTM混合模型在不同时间周期对不同类型的股票和股指均具有较好的预测效果。

Neural network model has always been an important method to solve the problem of quantitative investment in the financial field.In order to improve the accuracy of stock prediction and enhance the effectiveness and stability of stock price prediction,this study fuses convolutional neural network,LSTM model and wavelet denoising,and proposes WD-CNN-LSTM model.Firstly,the wavelet threshold denoising is used to filter and decompose the stock data to extract the low-frequency data,and then the processed data is extracted through the convolutional neural network for feature extraction.Finally,the LSTM model is used to process the extracted feature information to obtain the predicted stock price.The results show that the WD-CNN-LSTM hybrid model has good prediction effects on different types of stocks and stock indexes in different time frames.