Development of a Customer Churn Model for Banking Industry Based on Hard and Soft Data Fusion
Development of a Customer Churn Model for Banking Industry Based on Hard and Soft Data Fusion
复制标题
DOI：
10.1109/access.2023.3257352
复制
发表时间：
2023
期刊：
影响因子：
3.9
通讯作者：
Masoud Alizadeh;D. S. Zadeh;Behzad Moshiri;A. Montazeri
中科院分区：
文献类型：
--
作者：
Masoud Alizadeh;D. S. Zadeh;Behzad Moshiri;A. Montazeri
在过去的几年里，客户流失率有所增加--客户决定不再继续从组织购买产品或服务。客户的数据分为两类：软数据和硬数据。硬数据是指各种设备和程序产生的记录，包括但不限于智能手机、计算机、传感器、智能电表、车队管理系统、呼叫详细记录(CDR)和消费者银行交易数据。另一方面，受解释和观点影响的信息被称为“软数据”。融合这两种类型的数据可以更好地进行客户行为分析。本文使用一种有监督的机器学习算法，即决策树(DT)和变化挖掘方法对硬数据进行建模。K-均值聚类，一种无监督的机器学习算法，也与数据预处理技术一起使用。本文还考虑了Dempster-Shafer理论等软数据建模步骤。通过对软、硬数据的融合，可以计算出相互比较的客户流失率。此外，还利用客户的银行数据进行数据建模。研究结果表明，运用该模型将使银行业获得更具活力和效率的客户关系管理系统。
There has been an increase in customer churn over the past few years—customers decide not to continue purchasing products or services from an organization. Customers’ data lie in two categories: soft and hard. The term “hard data” refers to the records generated by various devices and programs, including but not limited to smartphones, computers, sensors, smart meters, fleet management systems, call detail records (CDRs), and consumer bank transaction data. On the other hand, information that is subject to interpretation and viewpoint is known as “soft data.” Fusing these two types of data leads to better customer’s behavior analysis. This paper uses a supervised machine learning algorithm, namely a decision tree (DT), and the change mining method to model hard data. K-means clustering, an unsupervised machine learning algorithm, is also used along with the data preprocessing techniques. This paper also considers the Dempster-Shafer theory and other steps for soft data modeling. By fusing soft and hard data, the churn rate of customers compared with each other can be calculated. Besides, the customers’ banking data are leveraged for data modeling. The results show that the banking industry will gain a more dynamic and efficient customer relationship management system by using this model.