基于分布式混合数据挖掘的电信客户流失分析

Telecommunication Carriers Customer Churn Analysis Based on Distributed Hybrid Data Mining

CORBA技术庞大而复杂,且技术和标准的更新相对较慢。电信运营企业应用系统是客户流失分析的主要数据来源,而传统的客户流失分析由于该系统数据的集中式存储继而采用集中式挖掘,对海量数据的挖掘效率低下。为进一步提高挖掘效率,提出网格下基于分布式混合数据挖掘的电信客户流失分析(Customer Churn Analysis upon Distributed HybridData Mining in Grid,CCA-DHDM),并借助GridSphere门户,在该平台上实现了BP神经网络算法和K-Means聚类算法。仿真实验表明,与单机环境相比,随着网格节点数增加,算法的平均耗时明显下降65%到75%,同时算法的效率得以较明显地提高。

CORBA is a large and complex technology,and the updating of technique and standard is relatively slow.Telecom enterprise application systems are the main source of data for customer churn analysis,the traditional customers churn analysis uses a centralized mining due to centralized data storage,and the mining efficiency for mass data is low.Present CCA-DHDM（Customer Churn Analysis upon Distributed Hybrid Data Mining in Grid）,and achieve the BP neural network algorithm and K-Means clustering algorithm in this platform by means of GridSphere Portal.The simulation shows that,compared with stand-alone environment,the average time-consuming of algorithm is decreased obviously 65 percent to 75 percent with the grid nodes increasing,and the efficiency of the algorithm is improved clearly.