摘要
客户流失管理是电信运营商通过对客户需求满意度调查进行有针对性挽留客户的一个重要方法,其中最关键的就是对客户流失行为做出预测。提出了一种基于神经网络的客户流失预测模型。根据行业专家经验值选取分析变量,通过神经网络计算分析变量的权值,建立客户流失预测模型并对客户流失趋势进行预测。该方法与决策树和贝叶斯网络等算法相比,通过使用两次神经网络,从原始数据上千个属性中提炼出与客户流失度相关性较大的属性,分析出的影响流失属性更利于下一步的客户挽留工作。
Churn management is a major focus of mobile operators to retain subscribers via satisfying their needs under resource constraints. One of the most important tasks is churn prediction. A customer churn prediction model based on neural network was put forward. Choose analysis variable according to expert empirical evaluation, and calculate the weight of analysis variable using neural network, set up customer churn prediction model and predict customer churn. Compared to decision tree and Bayesian network, the attributes closely related to churn degree were abstracted by using two neural networks. The attributes obtained are more favorable for customer detainment.
出处
《计算机应用》
CSCD
北大核心
2007年第9期2294-2297,共4页
journal of Computer Applications
基金
四川省科技厅科技攻关项目(2006Z01-011)
关键词
BP神经网络
关键绩效指标
客户流失
商业智能
Back Propagation Neural Network （BPNN）
Key Performance Indicator （KPI）
churn management
business intelligent
