Design and Implementation of an AI-Integrated Financial Decision Support System with LSTM and Random Forests

Abstract

Artificial intelligence is an application technology that affects the financial decision making process of enterprises and guides them to achieve intelligent financial operation and scientific management. This article designs an intelligent financial decision support system architecture that combines multiple functions and machine learning, rule based intelligent behavior, and human computer interaction, greatly enhancing the intelligent capabilities of enterprise financial data modeling and analysis, trend prediction, and implementation. The system integrates diversified data acquisition and cleaning, and completes data fusion preprocessing of financial, business, and external economic information; The financial data prediction model, which combines LSTM dominated time series models and random forest algorithms, enhances the real time prediction performance of cash flow, cost structure changes, and profit prospects; Implementing auxiliary decision making and adjustment optimization in complex environments through the integration of a knowledge base and rule base supported inference module. After testing various key performance indicators such as Financial Prediction Error (MAE), Reaction Time RT, and Benefit Improvement Rate COR, the performance is better than the current financial information system. After testing on Company A’s one year operational dataset, empirical evaluation shows that the proposed system reduces prediction error from 13.6% to 5.1% (a decrease of approximately 62.5%) and shortens average decision response time from 45s to 12s (a reduction of about 73%) compared with the baseline financial information system. This study provides a construction method and theoretical basis for an intelligent financial decision support system that is clearly understood, constantly growing, and self managed.