中文摘要：

目前客户流失预测面临的主要问题之一就是类不平衡性（class imbalance）。针对这个问题,首先应用欠抽样法（undersampling）处理客户流失数据降低不平衡性,再应用C4.5D、C4.5N、RIPPER、NaiveBayes和Ran-domForest机器学习方法对客户流失进行预测。实验结果表明,欠抽样法是在牺牲负类样本预测精度的前提下,提高正类预测精度,于是采用重复抽样法（resampling）来弥补欠抽样法的缺陷,减少负类样本中含有大量有用信息的丢失,实验结果证明了这种方法的正确性和有效性。

英文摘要：

Class imbalance is one of main obstacles in customer churn prediction. To overcome this problem,this paper applied undersampling technology to reducing class imbalance,and applied 5 classifiers,such as C4. 5D,C4. 5N,RIPPER,NaiveBayes and RandomForest,to predicting customer churn. Undersampling technology increased predictive accuracy of customer churn by sacrificing predictive accuracy of normal customer,according to experimental results. So,applied resample technology to preprocessing customer churn dataset. Proved resample technology to make up shortage of undersampling techno-logy. So,resample technology is more suitable for customer churn dataset,which have characteristic of class imbalance.