Abstract：

This study proposes an intelligent analysis scheme for complex networks based on Large Language Models (LLMs), aiming to address the issues of low efficiency, high professional threshold in traditional complex network analysis methods, and low accuracy when directly using large models. By constructing a multi-dimensional evaluation model and a Function Call mechanism, the scheme realizes the conversion of natural language instructions into background commands, automatically parses network data, invokes indicator functions, generates analysis reports, and integrates with the Neo4j graph database for visual display. Experiments show that this scheme significantly improves the automation level and accuracy of complex network analysis, reduces the usage threshold for non-professional users, and provides intelligent tool support for scenarios such as network security situation awareness.

Key Words：

complex network; Large Language Model (LLM); function call mechanism; key node identification; key link identification