基于改进Elman模型的电信公司客户流失分析  

Analysis of Customer Churn in Telecom Companies Based on Improved Elman Model

作　　者：曹宁 王雨薇 高莹[2] 徐根祺 任小文 CAO Ning;WANG Yu-wei;GAO Ying;XU Gen-qi;REN Xiao-wen(School of Civil Engineering,Xi'an Traffic Engineering Institute,Xi'an 710300,China;School of Civil Engineering,Xijing Universitx,Xi'an 710123,China;School of Electrical Engineering,Xi'an Traffic Engineering Institute,Xi'an 710300,China)

机构地区：[1]西安交通工程学院土木工程学院,西安710300 [2]西京学院土木工程学院,西安710123 [3]西安交通工程学院电气工程学院,西安710300

出　　处：《西安文理学院学报（自然科学版）》2022年第1期50-55,共6页Journal of Xi’an University(Natural Science Edition)

基　　金：陕西省教育厅专项科学研究计划项目(17JK1019);西安交通工程学院中青年基金项目(XJY212030)。

摘　　要：客户流失已成为目前企业管理中所面临的突出问题,对客户流失情况进行分析显得尤为重要,研究客户流失分析模型成为重中之重.为了解决传统客户流失预测方法准确率不高的问题,将Elman神经网络模型与灰色系统相结合对客户流失进行预测,利用X电信公司的客户样本数据进行仿真,同时将该模型和常用的遗传算法优化BP神经网络(GA-BP)模型、支持向量机(SVM)模型以及卷积神经网络(CNN)模型进行对比.结果表明,改进后的GM-Elman神经网络的精准率为85.21%、准确率为95.19%、F1值为77.33%、召回率为72.01%,均高于其它三种模型,改进Elman模型能够更高效地对客户流失进行预测.本研究为分析客户流失情况提供了一种新的方法.Customer churn has become a prominent problem in current enterprise management.It is particularly important to analyze customer churn,and the research on customer churn analysis model has become the top priority.In order to solve the problem of low accuracy of traditional customer churn prediction methods,Elman neural network model is combined with grey system to predict customer churn,and the customer sample data of X telecom company is used for simulation.At the same time,the model is compared with the commonly used genetic algorithm optimized BP neural network(GA-BP)model,support vector machine(SVM)model and convolutional neural network(CNN)model.The results show that the precisfon rate of the improved GM Elman neural network is 85.21%,the accuracy rate is 95.19%,the F1 value is 77.33%and the recall rate is 72.01%,which are higher than that of the other three models.The improved Elman model can predict customer churn more efficiently.This study provides a new method for analyzing customer churn.

关 键 词：灰色理论 神经网络 客户流失 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]