基于大小模型协同的智能化移动网络优化研究  

Research on Mobile Network Optimization Based on Collaboration of Large and Small Models

作　　者：黄金超 谢志普 吕非彼 狄子翔 邢震 程新洲[1] Huang Jinchao;Xie Zhipu;Lu Feibi;Di Zixiang;Xing Zhen;Cheng Xinzhou(China Unicom Research Institute,Beijing 100048,China)

机构地区：[1]中国联通研究院,北京100048

出　　处：《邮电设计技术》2024年第9期7-12,共6页Designing Techniques of Posts and Telecommunications

摘　　要：提出了一种大小模型协同的智能化移动网络优化方法,首先利用大语言模型处理和理解网络日志、外部开源等非结构化数据,从中提取关键数据。其次构建了一个包含网络设备、参数配置、专家优化经验等多维度信息融合的知识图谱,用于分析网络状态和优化需求之间的关系。然后,通过深度学习、图神经网络等专业工具模型进行根因分析,快速定位网络故障点,并基于专业知识图谱库和大模型的问题推理能力,辅助专家给一线员工提供具体的解决方案。最后,通过实际场景的实施和验证,由专家、一线员工对所提解决方案进行评估和反馈,这些评估和反馈信息经收集后不断返回,形成循环优化。It proposes an intelligent mobile network optimization method based on the collaboration of large and small models.Firstly,we utilize large language models to process and understand unstructured data such as network logs and external open source data,extracting key information.Secondly,a knowledge graph is constructed,encompassing multidimensional information such as network equipment,parameter configurations,and expert optimization experience,to analyze the relationship between network status and optimization requirements.Then,using specialized tools and models such as deep learning and graph neural networks for root cause analysis,network fault points can be quickly identified.Based on a knowledge graph database and the problem reasoning capabilities of large models,the proposed method can provide specific solutions to front-line employees.Finally,through the implementation and validation in real scenarios,experts and front-line employees evaluate and provide feedback on the proposed solutions.This evaluation and feedback information is continuously collected and fed back to the previous layers,forming a loop of continuous optimization.

关 键 词：移动网络优化 大小模型协同 知识图谱 图神经网络模型 大语言模型 

分 类 号：TN929.5[电子电信—通信与信息系统]