基于RFM与XGBOOST模型融合的应用研究——以客户流失预测为例  

Insurance customer churn prediction based on improved XGBOOST

作　　者：李佳镁 LI Jiamei(Jiangsu Branch office of China Post Company Group,Information Technology Center,Nanjing 210014,China)

机构地区：[1]中国邮政集团公司江苏省分公司信息技术中心,江苏南京210014

出　　处：《中国高新科技》2024年第19期54-57,共4页

摘　　要：文章提出将RFM模型客户价值分类结果作为特征融合机器学习算法的智能客户流失预测方法,并对其进行验证。首先,采用Pearson系数对特征进行相似度分析,以筛除冗余特征。其次,利用RFM价值分类模型,详细地对客户进行划分。最后,在XGBOOST算法中叠用RFM价值分析模型分析结果,同时结合其他特征维度分析客户流失情况,以增强判断精度。案例通过对中国国内某企业客户数据的预测结果验证发现,优化后的机器学习模型比常规模型的预测结果准确性明显提高。另外,经过改进后的XGBOOST模型,比修改之前的预估准确性、召回成功率分别提高了3.1%、2.3%。This study proposes an intelligent customer churn prediction method that combines the customer value classification results of the RFM model with machine learning algorithms,and validates it.Firstly,the Pearson coefficient is used to perform similarity analysis on data features and screen out redundant features.Secondly,the RFM value classification model is used to classify customers in detail.Finally,the result of RFM model will be overlaid in the XGBOOST algorithm,while combining other feature dimensions to analyze customer churn and enhance judgment accuracy.Through the verification of the prediction results of insurance customers data of a domestic enterprise in China,it was found that the optimized machine learning model significantly improved the accuracy of the prediction results compared to the conventional model.In addition,the improved XGBOOST model enhanced the accuracy of the prediction and recall success rate by 3.1%and 2.3%,respectively,compared to the model without modifications.

关 键 词：客户价值分类 流失预警 XGBOOST算法 RFM模型 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]