Application of SAO-Optimized Hybrid Neural Network Model in Stock Price Prediction

With the increasing complexity of global financial markets, traditional stock price prediction methods often perform poorly due to their inability to effectively handle nonlinear and high-dimensional data. Therefore, this paper constructs a hybrid SAO-CNN-BiLSTM-SA model, which uses convolutional neural networks (CNN) to automatically extract features from time series data, bi-directional long and short-term memory networks (BiLSTM) to capture temporal dependencies, self-attention self-attention mechanism (SA) to enhance understanding of global dependencies, and the SAO algorithm to optimize hyper parameters. The model is trained and tested using historical minute-level data from the Shanghai Stock Exchange Composite Index and compared with traditional CNN-BiLSTMSA models. Experimental results show that the SAO-CNN-BiLSTM-SA model significantly outperforms traditional models in key performance indicators such as root mean square error (RMSE), mean absolute error (MAE), and R-squared value. Especially in handling sudden market events and high volatility, the new model shows greater adaptability and accuracy. The successfully developed SAO-CNN-BiLSTM-SA model in this study not only improves stock price prediction accuracy but also demonstrates its application potential in complex financial data analysis by integrating the latest machine learning techniques and optimization algorithms.