基于多模态融合与图神经网络的用户精准感知系统研究  

Research on Customer’s Accurate Perception System Based on Multi-modal Fusion and Graph Neural Network Methods

作　　者：高伟[1] 汪悦 宋春涛[1] 程新洲[1] 王思维 孟范玉 Gao Wei;Wang Yue;Song Chuntao;Cheng Xinzhou;Wang Siwei;Meng Fanyu(China Unicom Research Institute,Beijing 100176,China;China United Network Communications Group Co.,Ltd.,Beijing 100033,China)

机构地区：[1]中国联通研究院,北京100176 [2]中国联合网络通信集团有限公司,北京100033

出　　处：《邮电设计技术》2023年第6期30-35,共6页Designing Techniques of Posts and Telecommunications

摘　　要：5G时代,通信运营商面临网络复杂化、业务差异化和用户需求多样化等挑战,引入人工智能技术,实现对用户的精准感知,并按需为用户提供个性化服务,已经成为运营商数字化转型的重点方向。研究了用户精准感知系统的构建方案,系统具备对运营商网络数据、业务数据以及声音、图像、文本数据进行多模态融合的能力,并能够通过图神经网络算法进行建模分析,实现对用户的精准感知与洞察。另外,介绍了基于用户精准感知系统进行高危流失用户挽留的应用场景,为人工智能技术在运营商数字化转型中的应用提供一种新的思路。In 5G era,communication operators face challenges such as complex network,diversified business and diversified customer needs.The introduction of artificial intelligence technology by operators to achieve accurate perception of customers and provide personalized services to customers on demand has become the key direction of digital transformation.It studies the construction scheme of the customer's accurate perception system.The system can realize multi-modal fusion of operator network data,business data,voice,image and text data,and use graph neural network algorithm to build models and accurately perceive and insight customers.In addition,it introduces a use case to reduce customer churn based on this system,and provides a new idea for the application of AI technology in the digital transformation of operators.

关 键 词：图神经网络 多模态融合 用户精准感知 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]