摘要
为了准确地识别高离网倾向的用户群,基于运营商的实际业务数据,通过运用数据挖掘中的CRISP-DM模型方法对移动用户的离网行为进行分析,并详细介绍了离网预测模型的建立过程。特别地,为了提高模型预测精度,将全量用户分为正常和非正常用户,对比分析这两类用户之间的差异并分别建模。经过数据实验证明了模型的有效性,能够达到实际商业应用的要求。
In order to accurately recognize users with high churn probability, we analyzed mobile user churn behavior using operators＇ actual service data and described the model building process in detail based on the CRISP-DM data mining method. Additionally, we distinguished users into normal and abnormal groups, compared the difference, and built models respectively in order to improve the prediction accuracy. The validity of the model was verified through data experiment, thus meeting the requirements of business application.
出处
《移动通信》
2016年第14期38-41,45,共5页
Mobile Communications
