Abstract
The stock market is a highly complex, dynamic, and nonlinear system, characterized not only by various types of random noise but also by the combined influence of macroeconomic policies, international environments, and investor behavior. Consequently, accurately predicting stock price movements remains a highly challenging task. Traditional single-stock prediction models often suffer from limited forecasting accuracy and struggle to capture nonlinear features effectively. To address these issues, this study constructs a hybrid CNN-LSTM prediction model based on historical data of the Shenzhen Component Index. This model leverages the feature extraction capabilities of convolutional neural networks (CNNs) and the temporal dependency modeling capabilities of long short-term memory networks (LSTMs) to enhance the fitting and predictive performance for complex stock price time series data. Experimental results demonstrate that compared to traditional methods, the CNN-LSTM hybrid model exhibits superior performance in both prediction accuracy and stability, achieving a coefficient of determination (R²) of 98.5%. These findings validate the feasibility and effectiveness of the proposed approach for stock price forecasting, providing not only a novel perspective for dynamic modeling of financial markets but also potential theoretical and practical value for financial decision-making and risk management.