Combined rough set theory and flow network graph to predict customer churn in credit card accounts
Introduction
Losing a customer is an opportunity for competitors to gain a customer. With so much competition, companies need to focus on keeping existing customers by satisfying their needs because the cost of attracting a new customer is usually considerably more than the cost to retain a current customer (Heskett et al., 1994, Reichheld and Sasser, 1990, Van den Poel and Lariviere, 2004). From the risk management perspective, retaining an existing customer lessens the need for a commercial bank to acquire a less credit-worthy customer or one whose ability or willingness to pay is uncertain. Customer characteristics become increasingly important in a competitive and mature credit card industry within which customers can easily switch their accounts and balances from one bank to another. In the past decade, with the help of business intelligence, databases have been growing rapidly. Commercial banks hold enormous amounts of their customers’ transaction data in customer relationship management (CRM) databases, including data related to sales, servicing and marketing functions. This data provides abundant information about customers that decision-makers can use to characterize customers for strategic planning and decision-making purposes and to enhance their competitiveness. However, data is only as good as the system that turns it into usable information. Given the proclivity of some customers to switch credit card companies, there is a pressing need for an integrated system that will identify the customer characteristics that lead to churn before the customers are lost so appropriate action can be taken (Chiang, Wang, Lee, & Lin, 2003). This kind of information can also be used to identify and target customers for implementation of customer management strategies to maximize customer value.
The original rough set theory (RST) proposed by Pawlak, 1982, Pawlak, 1984 is an effective approach for discovering hidden deterministic rules and associative patterns in all types of data and for handling unknown data distribution and information uncertainty or ambiguity. In other words, it can integrate learning-from-example technology, extract rules from data sets, and identify data regulations (Komorowski & Zytkow, 1997). Recently, RST has been applied in the marketing field because it is of benefit in analyzing and segmenting customer characteristics to formulate efficient and effective marketing strategies (Tseng & Huang, 2007), such as personal investment portfolio analysis (Shyng, Shieh, Tzeng, & Hsieh, 2009), video game customer purchase behavior (Tseng & Huang, 2007), insurance market attributes analysis (Shyng, Wang, Tzeng, & Wu, 2007), brand marketing (Beyon, Curry, & Morgan, 2001), supermarket customer loyalty (Lingras, Hogo, Snorek, & West, 2005) and travel demand analysis (Goh & Law, 2003).
Numerous studies have applied customer churn analysis to several areas, like the credit card industry (Kim et al., 2005, Kumar and Ravi, 2008, Lee et al., 2001, Lee et al., 2002), the wireless telecom industry (Hwang et al., 2004, Wei and Chiu, 2002) and the financial services industry (Van den Poel & Lariviere, 2004). Customer churn analysis benefits for managers in decisions about the right marketing strategy to retain their customers. However, RST has not been widely used in predicting customer churn, especially in credit card industry. Therefore, the major objective of this study is to adopt RST to predict the characteristics of credit card customers so decision-makers can understand the rules of customer churn and use multiple indicators to formulate efficient and effective strategies for marketing of reducing churn.
This article is organized as follows. Section 2 provides an overview of previous relevant studies in the domain of customer churn. In Section 3 we use the RST to inducing the rule of credit card customer churn. In Section 4, we design and develop a flow network graph based on the decision rules extracted by RST, and Section 5 concludes.
Section snippets
Review of credit card customer churn
Customer churn occurs when customers switch vendors or cancel service altogether and can be categorized as: unavoidable churn, involuntary churn and voluntary churn (Modisette, 1999). Unavoidable churn occurs when a customer dies or moves out the provider’s operating area. Involuntary churn occurs when a user fails to pay for service and the provider terminates services as a result. Termination of service resulting from theft, fraudulent service acquisition or fraudulent usage is also
Rough set theory and flow network graph
The rough set theory was first introduced by Pawlak, 1982. RST has been used by many researchers, and the theory has a long list of achievement (Pawlak & Skowron, 2007). This section reviews the basic concepts of rough sets and flow network graph.
An empirical credit card case
Based on the global financial tsunami that began in 2008, the number of viable credit card accounts in Taiwan is significantly decreasing, so customer churn is becoming an important issue for commercial banks. Commercial banks that understand the characteristics of customer churn are able to enhance customer loyalty and decrease churn. In this section, we use the ROSE2 (Rough Set Data Explorer) tool to adopt the rough set approach as analytical procedures, include: (1) selecting data and
Conclusion
This study uses a rough set approach and a flow network graph to predict credit card customer churn in a commercial bank in Taiwan. RST is used to discover hidden information in data and to exploring the rules and characteristics of customer churn. The decision rules can be transferred into a flow network graph to represent the connections of pathways and the degrees of their interdependency. Our empirical results, which illustrate customer characteristics as a predictor of churn forms, show
References (32)
- et al.
Separating financial from commercial customer churn: A modeling step towards resolving the conflict between the sales and credit department
Expert Systems with Applications
(2008) - et al.
Goal-oriented sequential pattern for network banking churn analysis
Expert Systems with Applications
(2003) - et al.
Toward a hybrid data mining model for customer retention
Knowledge-Based Systems
(2007) - et al.
Subscription fraud prevention in telecommunications using fuzzy rules and neural networks
Expert Systems with Applications
(2006) - et al.
Incorporating the rough sets theory into travel demand analysis
Tourism Management
(2003) - et al.
An LTV model and customer segmentation based on customer value: A case study on the wireless telecommunication industry
Expert Systems with Applications
(2004) - et al.
The state of CRM adoption by the financial services in the UK: An empirical investigation
Information & Management
(2005) - et al.
Temporal analysis of clusters of supermarket customers: Conventional versus interval set approach
Information Sciences
(2005) Rough classification
International Journal of Man–Machine Studies
(1984)Rough sets, decision algorithms and Bayes’ theory
European Journal of Operational Research
(2002)