ASKSSA-CNN-BiLSTM: A Novel Time Series Forecasting Model for Stock Price Prediction Based on An Enhanced Sparrow Search Algorithm

Stock price forecasting is highly challenging due to the nonlinear, noisy, and rapidly fluctuating characteris-tics of high-frequency financial time series. To address these issues, this study proposes ASKSSA-CNN-BiLSTM, a hybrid model that integrates an enhanced Sparrow Search Algorithm (ASKSSA) with a CNN-BiLSTM architecture. ASKSSA im-proves the original SSA by incorporating Good Nodes Set initialization, a Sine-Cosine producer update strategy, adaptive t-distribution perturbation, and a triangular-walk-based edge update mechanism, significantly strengthening global exploration and convergence stability. The model is evaluated using high-frequency CSI 300 stock data recorded every 10 minutes. Experimental results show that ASKSSA effectively optimizes the hyperparameters of the CNN-BiLSTM network, enabling themodel to better capture short-term fluctuations and bidirectional temporal patterns. Compared with CNN-BiLSTM, Adam-CNN-BiLSTM, and SSA-CNN-BiLSTM, the proposed ASKSSA-CNN-BiLSTM achieves the lowest RMSE (1.0512E-01) and the highest accuracy (94.0967%). These results demonstrate that ASKSSA-CNN-BiLSTM provides a robust and efficient solution for high-frequency stock price prediction and offers strong potential for broader time-series prediction applications.