基于联合神经网络的投诉预测模型研究  

Research on a complaint prediction model utilizing joint neural networks

作　　者：马晓亮 刘英[2,3] 高洁 MA Xiaoliang;LIU Ying;GAO Jie(Xidian University,Xi’an 710126,China;Guangzhou Branch of China Telecom Co.,Ltd.,Guangzhou 510620,China;Ma Xiaoliang Innovation Studio for Model Workers and Creative Talents,Guangzhou 510620,China)

机构地区：[1]西安电子科技大学,陕西西安710071 [2]中国电信股份有限公司广州分公司,广东广州510620 [3]马晓亮劳模与工匠人才创新工作室,广东广州510620

出　　处：《电信科学》2024年第1期48-58,共11页Telecommunications Science

摘　　要：对影响电信运营商重复投诉的关键因素进行深入探讨,旨在提高服务质量并构建风险预测模型。基于运营商客服数据,研究采用了Logistic回归、BP神经网络以及二者联合建模的方法。Logistic回归模型确定了5个主要影响因素,预测重复投诉发生的概率,精度达到80.0%。BP神经网络则选取了81个影响因素,预测精度为90.6%。在此基础上,构建了联合模型,其精度高达92.8%。实际应用于某省会电信运营商后,重复投诉率下降了3.2%,成效显著,为提高电信运营商服务质量、降低重复投诉率提供了有力支持,对我国电信行业发展具有重要意义。By conducting in-depth exploration on the key factors affecting repeat complaints of telecom operators,this study aimed to improve service quality and construct a risk prediction model.Based on the operator’s customer service data,the study employed Logistic regression,BP neural network,and their combined modeling methods.The Logistic regression model identified five major influencing factors,predicting the probability of repeat complaints with an accuracy of 80.0%.The BP neural network selected 81 influencing factors,achieving a prediction accuracy of 90.6%.On this basis,a combined model was constructed with an accuracy rate of up to 92.8%.After practical appli-cation in a provincial telecom operator,the repeat complaint rate decreased by 3.2%,demonstrating a significant im-pact.Strong support is provided for improving the service quality of telecom operators and reducing repeat com-plaints,which is of great significance for the development of the telecom industry in China.

关 键 词：AI客服 联合建模 重复投诉 LOGISTIC回归 深度学习模型 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]