基于神经网络算法的电信客户流失预测  

Telecom Customer Churn Prediction Based on Neural Network Algorithm

作　　者：于明鑫 郑雅匀 YU Mingxin;ZHENG Yayun(Chengdu College of University of Electronic Science And Technology of China,Chengdu 611731,China)

机构地区：[1]电子科技大学成都学院,四川成都611731

出　　处：《现代信息科技》2023年第2期30-33,共4页Modern Information Technology

摘　　要：电信客户流失预测能够有效地帮助运营商制定有用的挽留策略。文章收录了来自某开源数据平台的电信公司数据集,该数据集包含了22个字段,20个特征变量。文章通过SPSSModeler对数据集进行分析,通过比较贝叶斯网络、神经网络算法和C5.0算法的预测准确率、ROC值以及ROC曲线下面积,最终发现神经网络算法通过训练2次后的效果较好。Telecom customer churn prediction can effectively help operators develop useful retention strategies.This paper includes a data set of telecom companies from an open source data platform,which contains 22 fields and 20 characteristic variables.This paper analyzes the data set through SPSSModeler,and compares the prediction accuracy,ROC value,and area under the ROC curve of Bayesian network,neural network algorithm,and C5.0 algorithm.Finally,it is found that the neural network algorithm has a better effect after twice training.

关 键 词：C5.0 贝叶斯网络 电信客户流失预测 神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]