基于PMML的移动通信客户流失模型交换

Churn Model Exchange in Mobile Communication Based on PMML

移动通信领域迫切需要在地理分布的经营分析系统之间交换标准的数据挖掘模型。尽管预测模型标记语言已经成为数据挖掘模型交换格式的业界标准,但并没形成可用的框架来指导标准交换模型的生产过程。该文提出了支持挖掘模型交换和移动通信客户流失分析的决策树算法框架。利用该框架构建了流失预警系统,并使用模拟客户数据验证了其有效性。对标准交换模型进行了适当扩展,以支持对移动通信数据更加有效的流失分析。

The mobile communication industry witnesses an increasing demand to share and exchange standardized data mining models among geographically distributed business analysis systems. Although PMML is recognized as the industrial standard for mining model exchange format, no framework has been developed to guide the production process of standard exchange models. This paper proposes a framework for decision tree construction algorithms that supports both model exchange and mobile communication churn analysis. A churn predictor conforming to this framework algorithm is implemented, and its validity is testified with simulated customer churn records. Proper extension is made to the standard exchange model to support more effective churn analysis of mobile communication data.