Exploring Network-Knowledge Graph Duality: A Case Study in Agentic Supply Chain Risk Analysis

2025-10-01

AI Summary

Large language models (LLMs) struggle with complex, multimodal, and network-native data prevalent in financial risk analysis—particularly in supply chain contexts. Method: We propose a lightweight agent framework that models supply chain networks as knowledge graphs, leveraging their duality with domain knowledge graphs; integrates numerical factor tables, news streams, and network topology; employs a “context shell” template for lossless numerical-to-natural-language conversion; and combines centrality-driven graph traversal with retrieval-augmented generation to automatically identify critical risk pathways and generate real-time, interpretable risk narratives. Contribution/Results: The method requires no LLM fine-tuning and operates without specialized graph databases. It achieves high efficiency and strong interpretability while maintaining low computational overhead, significantly improving both the timeliness and decision-support capability of supply chain risk analysis.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Natural Language Processing: (Large) Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Large Language Models (LLMs) struggle with the complex, multi-modal, and network-native data underlying financial risk. Standard Retrieval-Augmented Generation (RAG) oversimplifies relationships, while specialist models are costly and static. We address this gap with an LLM-centric agent framework for supply chain risk analysis. Our core contribution is to exploit the inherent duality between networks and knowledge graphs (KG). We treat the supply chain network as a KG, allowing us to use structural network science principles for retrieval. A graph traverser, guided by network centrality scores, efficiently extracts the most economically salient risk paths. An agentic architecture orchestrates this graph retrieval alongside data from numerical factor tables and news streams. Crucially, it employs novel ``context shells''-- descriptive templates that embed raw figures in natural language -- to make quantitative data fully intelligible to the LLM. This lightweight approach enables the model to generate concise, explainable, and context-rich risk narratives in real-time without costly fine-tuning or a dedicated graph database.

Problem

• Addressing LLMs' limitations with complex financial risk data
• Exploiting network-knowledge graph duality for supply chain analysis
• Enabling real-time risk narratives without costly fine-tuning

Innovation

• Agent framework exploits network-knowledge graph duality
• Graph traverser uses centrality scores for risk path extraction
• Context shells embed quantitative data in natural language templates