当前,银行产品同质化,客户选择产品和服务的方式越来越多,客户对产品的忠诚度越来越低,客户流失成为银行业最关注的问题之一。获得新客户的成本远远高于维持老客户的成本。
[En]
At present, there is homogenization of banking products, there are more and more ways for customers to choose products and services, and customers' loyalty to products is getting lower and lower, so customer churn has become one of the most concerned issues in the banking industry. The cost of getting new customers is much higher than the cost of maintaining old customers.
研究表明,商业银行客户流失较为严重。在国内商业银行,客户流失率可以达到20%甚至更高。获得新客户的成本可能是维持现有客户成本的5倍。因此,从大量的客户交易记录中挖掘出对客户流失有影响的信息,建立高效的客户流失预警系统就显得尤为重要。
[En]
According to the research, the customer loss of commercial banks is more serious. In domestic commercial banks, the customer turnover rate can reach 20% or more. The cost of obtaining new customers can be up to 5 times that of maintaining existing customers. Therefore, it is particularly important to dig out the information that has an impact on customer churn from a large number of customer transaction records and establish an efficient early warning system for customer churn.
下面小编就以某商业银行客户流失预测案例,帮助大家了解数据分析的过程。以下数据分析过程依托国内排名靠前的BI软件Smartbi一站式数据分析平台,可以大大降低数据分析的复杂度,打破数据孤岛的困境,并且通过简单鼠标拖拉拽的方式就可以快速轻松完成数据可视化分析,让企业都在第一时间了解到业务数据指标的变化情况。
首先,我们需要分析当前的业务形势,选取近一年的高价值零售客户群体进行分析。如果我们发现离职率非常严重,我们需要建立针对高价值客户群体的流失预警模型。摸清客户流失原因,引导业务加强客户维护,增强客户对银行产品的粘性。
[En]
First of all, we need to analyze the current business situation and select the high-value customer group of retail customers for nearly a year to analyze. If we find that the turnover rate is very serious, we need to establish a loss early warning model for high-value customer groups. Find out the reasons for customer loss, guide business to strengthen customer maintenance, and enhance customer viscosity to the bank's products.
该分析思路可以从高价值客户群体的个人信息、账户信息、交易信息等维度数据入手,结合第三方数据,利用随机森林算法构建客户流失预警模型,输出影响客户流失的主要因素。
[En]
The analysis idea can start with the dimensional data such as personal information, account information and transaction information of the high-value customer group, as well as combine the third-party data, use the random forest algorithm to construct the customer churn early warning model, and output the main factors that affect the customer churn.
数据来源:
数据来源于CRM系统中客户基本信息表、账单表等;第三方数据,数据时间窗为近一年的数据,客群为高价值客群,本次案例已获取到部分数据总共100000条数据。
数据维度信息包含:
银行自有字段:账户类信息、个人类信息、存款类信息、消费、交易类信息、理财、基金类信息、柜台服务、网银类信息;
外部三方数据:外呼客服数据、资产类数据、其他消费类数据;
本次案例流失定义为:3个月内没有与银行业务任何往来的客户。
整个数据预处理流程图:
我们通过相关性节点将各特征指标数据进行 相关性分析,方便特征选择进入模型训练,如图:
通过分析发现:是否代发客户、卡等级、月均代发金额、最多代发金额、月均AUM、月初AUM与是否流失都具有相关性,其他特征与是否流失相关性为0。
因此,无论Label列是否丢失,我们都通过特征来选择相关的特征,如图所示。
[En]
Therefore, we select the relevant features through the features, as shown in the figure, whether the label column is lost or not.
模型训练
在这种情况下,随机森林算法被采样用于模型训练,并通过节点分裂将数据按7:3的比例划分为训练集和验证集。
[En]
In this case, the random forest algorithm is sampled for model training, and the data is divided into training set and verification set according to proportion 7:3 by splitting nodes.
整个模型训练流程如图所示:
参数配置如图:
模型评估
我们通过评估节点对数据进行评估,如模型培训流程图所示,评估结果如图所示:
[En]
We evaluate the data through the evaluation node, as shown in the model training flow chart, and the evaluation results are as shown in the figure:
我们发现评估结果中F1得分为0.95,说明模型预测的效果比较好的。
业务分析
我们通过随机森林特征选择节点输出重要性较高的5个特征,结果如图:
通过对某业务线条高价值客群进行流失预警分析,发现影响客户流失的主要因素为:月均AUM、月初AUM、卡等级等。主要原因可能为产品缺乏竞争力、活动较少等。
因此,我们可以采取相关措施和建议,如:加强客户关系维护、产品跟踪、维护走访、跟踪系统、扩大销售、机制维护等。
[En]
Therefore, we can take relevant measures and suggestions, such as: strengthen customer relationship maintenance, product follow-up, maintenance visit, tracking system, expand sales, mechanism maintenance and so on.
Original: https://blog.csdn.net/Moogical/article/details/123201127
Author: 思迈特Smartbi
Title: Smartbi助你解决银行高价值客户流失难题

用于图像描述的评价指标CIDEr讲解

YOLOv5简析

Tensorflow(三) tf.keras的相关东西

GPU的使用说明

TensorFlow、CUDA、cuDNN版本对应关系

树莓派——槑槑智能音箱

深度学习 | 适配tensorflow2.6的CUDA与cuDNN

GPU云服务器平台对比!哪家最值得推荐?

语音模块 STC11L08XE代码功能简介

YOLOv5输入端(一)—— Mosaic数据增强|CSDN创作打卡

使用Pytorch中的tensorboard可视化网络训练参数

kaldi 的常用ark scp命令

【简单明了】anaconda 安装 jupyter 创建 tensorflow-gpu环境,深度学习模型训练用gpu跑,jupyter和pycharm上运行成功

免费教程·开源 | 从零开始制作ROS无人竞速车RACECAR教程
