一种基于Fisher比率和预测风险准则的电信客户流失预测分步特征选择方法

发布时间：2018-05-26 02:36

  本文选题：大数据 + 流失预测　； 参考：《中国科学技术大学学报》2017年08期

【摘要】：电信客户流失预测是电信运营商客户关系管理系统的一个重要问题,其目的是预测具有较高流失风险的客户.电信客户流失预测模型的构建过程包括数据预处理、不均衡处理、特征选择和分类器的训练与评估.针对电信数据集中存在的特征维度过高问题,结合过滤式特征选择和嵌入式特征选择方法的优点,提出了一种基于Fisher比率和预测风险准则的分步特征提取方法.结合真实数据集的实验结果表明,该方法能够减少特征维度,提高分类器的预测效果.
[Abstract]:Customer churn prediction is an important problem in telecom operators' customer relationship management system, and its purpose is to predict customers with high risk of loss. The construction process of telecom customer churn prediction model includes data preprocessing, unbalanced processing, feature selection and classifier training and evaluation. Aiming at the problem of high feature dimension in telecom data set, a step feature extraction method based on Fisher ratio and predictive risk criterion is proposed, which combines the advantages of filtering feature selection and embedded feature selection. The experimental results of real data sets show that this method can reduce the feature dimension and improve the prediction effect of classifier.
【作者单位】： 中国科学技术大学自动化系;
【基金】：Supported by the National Natural Science Foundation of China(61375079)
【分类号】：F626
，

本文编号：1935618