In this study, a multi-dimensional prediction model and visual decision support system integrating macroeconomic indicators are constructed for the dynamic monitoring needs of enterprise financial management. It achieves ±7% prediction accuracy improvement (53% error reduction compared with traditional model) by introducing Bayesian LSTM network, and develops interactive causal inference engine to support dynamic scene simulation. Empirical evidence shows that the system can reduce the rolling loan cost of manufacturing enterprises by 23% and the risk of inventory backlog in retail industry by 3.2 million RMB. The research innovations are: 1. proposing a multimodal data-driven financial health assessment system 2. establishing a flexible budget derivation algorithm that integrates policy text analysis.
Chenxi Fan (Thu,) studied this question.