基于PrefixSpan和LightGBM的网元拓扑连接关系判别方法  

A method for Identify Network Element Topology Connection Relationships Based on PrefixSpan and LightGBM

作　　者：倪晋宇 涂泾伦 杨天昊 陈晓峰 白云飞 NI Jinyu;TU Jinglun;YANG Tianhao;CHEN Xiaofeng;BAI Yunfei(China Mobile Communications Group Anhui Co.,Ltd.,Hefei 230000,China;China Mobile Communications Corporation,Beijing 100033,China;Bright Oceans Inter-Telecom Beijing Co.,Ltd.,Beijing 100043,China)

机构地区：[1]中国移动通信集团安徽有限公司,安徽合肥230000 [2]中国移动通信集团有限公司,北京100033 [3]亿阳信通股份有限公司北京分公司,北京100043

出　　处：《数字通信世界》2024年第1期41-44,89,共5页Digital Communication World

摘　　要：文章创新地提出了一种基于PrefixSpan和LightGBM的网元拓扑连接关系判别的方法,采用PrefixSpan算法对告警数据进行抽取挖掘,然后将挖掘结果进行分析并将分析结果输入到LightGBM中进行监督学习,获得最终网元拓扑连接关系判定模型。实验结果表明:本方法在基站及相关网元拓扑连接关系的推断中f1值达到了0.89,有效提升了网元拓扑连接关系判别的准确度,为网元拓扑连接关系校正提供了有力手段,为数字孪生网络构建打下坚实的基础。This article innovatively proposes a method for network element topology connection relationship discrimination based on PrefixSpan and LightGBM.The PrefixSpan algorithm is used to extract and mine alarm data,and the mining results are analyzed.The analysis results are then input into LightGBM for supervised learning to obtain the final network element topology connection relationship judgment model.The experimental results show that the f1 value of this method in inferring the topological connection relationship between base stations and related network elements reaches 0.89,effectively improving the accuracy of network element topological connection relationship discrimination,providing a powerful means for correcting network element topological connection relationships,and laying a solid foundation for the construction of digital twin networks.

关 键 词：数字孪生网络 频繁项集 时序 网元拓扑连接 机器学习 

分 类 号：TN929.52[电子电信—通信与信息系统]