摘要
用户流失问题是电信运营商面临的亟待解决的问题,针对不同的场景,业界研究开发了多个用户离网预测系统。服务号码捆绑指用户在使用运营商服务期间,与银行、电商、便利店等第三方服务提供商通过绑定手机号产生联系。通过研究发现用户在服务存续期间普遍会绑定多种第三方服务提供商,这些商家会不定时给用户推送短信,当用户即将流失时,多数用户会逐渐取消这类服务的绑定。因此,服务号码捆绑特征对于离网用户的甄别起到了重要的作用。采用随机森林算法构建离网预测模型,利用逻辑回归算法对服务号码捆绑特征进行降维,并加入模型,进行离网用户分析,从而辅助决策者制订相应的客户维挽策略,降低客户离网率。实验结果表明,服务号码软捆绑特征能够提高系统的分析预测能力。
Customer churn problem is one of the biggest problems that telco operators are faced with,and it need to be solved.For different scenarios,operators have developed many churn prediction systems.Service number bundling means that customers have relations with 3rd-service support companies such as bank,E-commercials,supermarkets etc.during service available period through bundling phone number.After researching,we found that customers will have service number bundling behaviors with kinds of 3rd-service support companies during service available period.These companies will send SMSs to customers at any time.For customers who are about to churn,they will cancel the service bundling gradually.Thus,service bundling feature plays an important role in predicting potential churners.In order to help decision-making manager formulate corresponding customer retention campaigns and drop the churn probability,we developed the churn prediction model with random forest algorithm,and used logistic regression algorithm to reduce service number features.Experiment results show that service number soft bundling can improve the analysis and prediction performance of churn prediction system.
出处
《计算机科学》
CSCD
北大核心
2016年第S2期585-590,共6页
Computer Science
基金
江苏省科技支撑计划重点项目(BE2014005-4)资助
关键词
客户流失
随机森林
服务号码捆绑
逻辑回归
离网预测系统
Customer churn
Random forest
Service bundling
Logistic regression
Churn prediction system
