数据挖掘在电信行业客户流失预测中的应用

The application of data mining to client churning prediction in telecom

客户流失是电信行业发展过程中所面临的一个严重问题,直接影响到运营商的企业效益。本文主要介绍了对电信行业客户流失情况进行数据挖掘的过程,改进了已有模型存在的缺乏灵活性、难以处理高维度数据的缺点,根据运营商的历史数据资料,利用SAS/EM模块对客户的固有特征和行为特征进行挖掘分析,采用决策树分类算法的CART算法建立了聚类分析模型和包括评估模块在内的一套完整的流失预测模型,能够直观地显示出流失客户的基本特征,并且可以对任意的数据集进行分析,有效提高了模型的普遍应用性和准确性。

Client churning is a serious problem in the development of telecommunication industry, and it has immediate influence to the profit of a company. This paper mainly introduces the whole procession of data mining in client churning of telecommunication. According to the data in the provider′ s database, by analyzing and mining the natural attribution and action attribution among the clients, we set up a clustering model and an integrated prediction model, including assessment module, which is based on CART algorithm of decision tree in SAS EM module for client churning. The new model improves the disadvantages of the existed models, such as lack of flexibility, unable to process data with high dimensionality, even shows the essential features of customers lost visually. Using this model can analyse arbitrary datasets effectively and it enhances the generational applicability and the prediction accuracy rate.