Scholarlink

<p>Time series data refers to a series of observation values arranged in a time sequence, which is widely used in the financial field. However, due to the complexity, variability, nonlinearity, and seasonality of time series data, the traditional time series prediction model cannot adequately address the nonlinearity in the data, and the single neural network model has limitations in feature extraction and parameter optimization. In this work, a new GWO-CNN-SE-LSTM model is proposed to predict financial time series data. Based on the Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) model, the Grey Wolf Optimization Algorithm (GWO) is introduced to optimize the model parameters, and the SE-attention mechanism is incorporated to enhance the network’s feature extraction capability. In this paper, RMSE, <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^{2}\)</EquationSource> </InlineEquation>, MAE, and MAPE are selected as evaluation metrics for the model. By using three sets of single-feature time series data from Consumer Price Index (CPI), Shanghai Stock Exchange Index (SSEI), and Producer Price Index (PPI), the GWO-CNN-SE-LSTM model improves the prediction accuracy of these three datasets by 60.08%, 18.47%, and 69.26%, respectively, compared with the LSTM model. When compared with the CNN-LSTM model, the improvements are 39.38%, 16.43% and 69.09%, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Single-Feature Financial Time Series Forecasting Model Based on CNN-LSTM with SE-Attention Mechanism and Grey Wolf Optimization Algorithm

  • 期刊论文

  • 发表日期：

    2025年12月16日

查看全文
  • Qian-Qian Zhang,
  • Min Li,
  • Tao Xu,
  • Shao-Qun Dong

摘要

Time series data refers to a series of observation values arranged in a time sequence, which is widely used in the financial field. However, due to the complexity, variability, nonlinearity, and seasonality of time series data, the traditional time series prediction model cannot adequately address the nonlinearity in the data, and the single neural network model has limitations in feature extraction and parameter optimization. In this work, a new GWO-CNN-SE-LSTM model is proposed to predict financial time series data. Based on the Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) model, the Grey Wolf Optimization Algorithm (GWO) is introduced to optimize the model parameters, and the SE-attention mechanism is incorporated to enhance the network’s feature extraction capability. In this paper, RMSE, \(R^{2}\) , MAE, and MAPE are selected as evaluation metrics for the model. By using three sets of single-feature time series data from Consumer Price Index (CPI), Shanghai Stock Exchange Index (SSEI), and Producer Price Index (PPI), the GWO-CNN-SE-LSTM model improves the prediction accuracy of these three datasets by 60.08%, 18.47%, and 69.26%, respectively, compared with the LSTM model. When compared with the CNN-LSTM model, the improvements are 39.38%, 16.43% and 69.09%, respectively.