Financial Management Early Warning System based on Improved Bi-Directional Gated Recurrent Unit
Yao Xia · 2025
In recent years, predicting financial risk warnings is significant for protecting economic stability, and enabling realistic actions over market crises, systemic risks and credit defaults. Previous researchers suggested various traditional models but still, because of the complexity of and high-dimensional financial data there are challenges such as; noise, macroeconomic factors and thus facilitates model interpretability for stakeholders, and computing uncertainty to improve confidence in predictions. In this research, Improved Bi-Directional Gated Recurrent Unit (Bi-GRU) method is proposed for effective and early financial risk prediction. Initially, Factor Analysis (FA) is used for acquiring common factors among the non-financial and financial indicators. Further, Improved Bi-GRU is employed for predicting the financial risks by capturing complex relationships and focuses on relevant information within the financial data can lead to effective predictions of market trends, stock prices and other financial indicators. Then, the Particle Swarm Optimization (PSO) is used to optimizes learning rates of Improved Bi-GRU. From the results, the Improved Bi-GRU model attains effective results including MSE (0.0029), MAPE (2.1987) and MAE (0.0489) respectively compared to the FA, Particle Swarm Optimizer and Long Short-Term Memory namely FA-PSO-LSTM.