Sparrow Search Algorithm-Optimized Long Short-Term Memory Model for Stock Trend Prediction.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Financial Forecasting

Background:

• Long Short-Term Memory (LSTM) networks are effective for time series analysis, including stock market prediction.
• Traditional hyperparameter selection for LSTMs relies on subjective experience, limiting model generalization.
• Optimal hyperparameter tuning is crucial for accurate stock trend prediction models.

Purpose of the Study:

• To propose an optimized LSTM model using the Sparrow Search Algorithm (SSA) for improved stock trend prediction.
• To enhance the generalization capability and accuracy of stock forecasting models.
• To address the limitations of subjective hyperparameter selection in LSTM networks.

Main Methods:

• Development of a Sparrow Search Algorithm-optimized LSTM (SSA-LSTM) model.
• Utilizing SSA to identify optimal hyperparameters for the LSTM network.
• Applying the SSA-LSTM model to Shanghai Composite Index stock data for trend prediction.

Main Results:

• The SSA-LSTM model demonstrated high forecasting precision on Shanghai Composite Index data.
• Achieved a Mean Absolute Percentage Error of 0.0093 and a Coefficient of Determination of 0.9754.
• Outperformed traditional stock forecasting methods in accuracy and interpretability.

Conclusions:

• The SSA-LSTM model offers a superior approach to stock trend prediction compared to traditional methods.
• Optimizing LSTM hyperparameters with SSA significantly enhances predictive accuracy and model generalization.
• The proposed model provides a more interpretable and accurate tool for financial market analysis.