Information-based Decision Support System Driven by Artificial Intelligence

Background: Traditional decision support systems face challenges in multi-source heterogeneous data fusion, dynamic scenario adaptation, and model interpretability, especially in time-sensitive domains like emergency management. Method: This study proposes an AI-driven dynamic cognitive decision-making framework. It integrates Graph Attention Networks (GAT) for multimodal data association, a hybrid Temporal Convolutional Network (TCN)-Transformer architecture for spatiotemporal feature extraction, and Grad-CAM for decision path visualization. A reinforcement learning mechanism with multi-agent co-evolution optimizes dynamic strategies. Results: Experiments demonstrate significant improvements: (1) Data fusion efficiency (5.5s for 10,000 entries vs. 7.2s/9.8s in baselines); (2) Decision optimization (92% material dispatch pass rate vs. 75%/80%); (3) Enhanced interpretability (75% decision path transparency score vs. 70%/72%). Conclusion: The framework effectively addresses efficiency, adaptability, and transparency limitations. Future work will optimize computational scalability and extend applications to domains like smart transportation.