FD-GRNet: A Dendritic-Driven GRU Framework for Advanced Stock Market Prediction

Abstract

Time series forecasting in financial markets presents significant challenges due to the inherent nonlinearity, volatility, and dynamic nature of market data. The unique architecture of the flexible dendritic-driven gated recurrent network (FD-GRNet) enables it to effectively manage both long-term dependencies and nonlinear patterns in financial time series. In pursuit of this capability, FD-GRNet integrates two novel components: the flexible dendritic neuron model (FDNM), enhancing the model's capacity to capture complex nonlinearities, and the dendritic gated recurrent network (DGRNet), improving its ability to handle temporal dependencies. Comprehensive experiments on major global stock market indices demonstrate that FD-GRNet consistently outperforms several comparative algorithms across multiple evaluation metrics. Ablation studies further highlight the essential roles of FDNM and DGRNet in improving the model's accuracy and robustness. Future research will focus on optimizing the model for broader time series forecasting applications.