Data Mining Application in Customer Relationship Management Of Credit Card Business Ruey-Shun Chen, Ruey-Chyi Wu and J. Y. Chen Institute of Information Management, National Chiao Tung University, Taiwan.
[email protected] business asset and aims to retain customers and enhance customer satisfaction. According to Caldwell [3], CRM is not a new concept. Many businesses have practiced it for a long time, for instance, by memorizing customer’s background and spending habits and introducing promotions targeting certain customers based on the information obtained.
Abstract First, we classify the selected customers into clusters using RFM model to identify high-profit, gold customers. Subsequently, we carry out data mining using association rules algorithm. We measure the similarity, difference and modified difference of mined association rules based on three rules, i.e. Emerging Patten Rule, Unexpected Change Rule, and Added/Perished Rule. In the meantime, we use rule matching threshold to derive all types of rules and explore the rules with significant change based on the degree of change measured. In this paper, we employ data mining tools and effectively discover the current spending pattern of customers and trends of behavioral change, which will allow management to detect in a large database potential changes of customer preference, and provide as early as possible products and services desired by the customers to expand the clientele base and prevent customer attrition. Keywords: Data mining, Customer relationship management, Credit card.
2.2. Experimental Model Bult and Wansbeek [4] point out that RFM (recency, frequency and monetary value) is most commonly used for selection or segmentation analysis in direct marketing; through RFM, marketer can sort out target customers from a huge list of customers for its marketing activity. Sung and Sang [5] use non-transformed RFM values directly as input variable for model building and then categorize customers into groups using cluster analysis where different marketing strategies may be formulated for different customer clusters. Goddman [6] holds that RFM analysis helps avoid waste of time and energy on cultivating low-profit customers and generate better return on capital by investing more marketing resources on higher profit customer segments. The customer value matrix proposed by Marcus [7] is a method suitable for small and med-sized enterprises to analyze customer value. Customer value matrix uses two variablesfrequency and money to explain customer value, while the third variable “recency” provides the time of customer contact to combine with frequency and average monetary value of purchase. Figure 1 depicts the customer value matrix.
1. Introduction Customer relationship management, through which, banks hope to identify the preference of different customer groups, products and services tailored to their liking to enhance the cohesion between credit card customers and the bank, has become a topic of great interest. Shaw et al. [9] mainly describes how to incorporate data mining into the framework of marketing knowledge management. The systemic application of data mining techniques reinforces the knowledge management process and allows marketing personnel to know their customers well to provide better services. Differing from Berry and Linoff [10], Shaw proposes the view of data visualization, believing that data visualization software allows marketers to see complex data depicted in three dimensions or colors and to slice, rotate or zoom the data to obtain different levels of details. Song [8] depicts a method to detect changes of customer behavior at different time snapshots from customer profiles and sales data. The common approach is to discover changes from two datasets and generate rules from each dataset to carry out rule matching. In this study, we propose three types of behavioral change rules—Emerging Pattern Rule, Unexpected Change Rule, and Added/Perished Rule.
Figure 1 Customer value Figure 2 System Framework Matrix Data mining helps users to identify valuable patterns contained in diverse data and their relations so as to help the major decisions. Currently businesses apply data mining in many fields, i.e finance, banking, and manufacturing. Generally speaking, data mining can create the following models: (1) classification; (2) cluster analysis; and (3) association analysis.
2. Literature review 2.1. Customer relationship management Kalakota & Robinson [1] defines CRM as the strategy integrating sales, marketing and service, which unites operating procedures and technology to better understand customers from different perspectives. Kandell [2] views CRM as a customer centric initiative that regards customer lifecycle as an important
3. Problem description Using the association rules in data mining techniques to analyze in detail customer transactions can learn which products might be purchased by customers at the same time, and based on the rules for combining popular products,
1
Proceedings of the 29th Annual International Computer Software and Applications Conference (COMPSAC’05) 0730-3157/05 $20.00 © 2005 IEEE
understand customer needs and adjust their marketing strategy accordingly to seek maximum return under limited resources.
marketing personnel or corporate decision makers can formulate more appealing marketing plan or operational rules and actively offer products that might interest the customers. Thus clusters may be used as basis for segmenting marketing targets. In consideration that customer’s spending behavior might change with time, we will also explore such time-varying and dynamically adjusted spending patterns.
6.1. Classification analysis We clustered the 2003 customers using the RFM model. Given that “most recent purchase date (Recent)” bears no significant influence in the analysis of customer value in credit card transactions, we considered only two factors -frequency and monetary value (M) in clustering analysis. Cluster 0 and Cluster 2, which collectively accounted for 80% of bank customers, but had brought little profit to the bank (spending less than $4500,000 per person each year), banks are advised against spending too much energy and costs on these types of customers. Cluster 3 and Cluster 4 were more loyal, but did not generate high profit for the bank. How to enhance their value to the bank should be the focus of marketing efforts. Cluster 1 accounted for 3.3% of credit card customers and they were the gold customers with high loyalty and high profit, and hence one of the bank’s most important assets. Losing such customers will result in substantial loss to the bank. Thus the bank should endeavor to provide better services and greater added value to retain them. A business only stands to win if it is able to use information and professional knowledge effectively to create value for its customers and solidify its relationship with them. At 5.5% support and 75% confidence level, we identified 641 association rules in 2003 dataset and 578 association rules in 2004 dataset using the Apriori algorithm provided in Weka.
4. System analysis and design 4.1. System framework As shown in Figure 2, the system framework for the study contains customer clustering, creation of association rules, and mining change of rules at different time.
4.2. Relationship between rules at different time
periods We identify higher profit customers using clustering algorithm, select datasets of two different time periods, and generate the association rules in those two periods using association rules algorithm. We first measure the similarity and difference between the rules, then use rule matching threshold to decide the type of the rule, and finally evaluate the degree of change to identify a significant trend of spending behavioral change. We use the term “similar” and “different” to depict the disparity between rules. Here we use a rule matching threshold (RMT) [8] to decide the measure of difference between rules, shown in Figure 3.
7. Conclusion We used association rules algorithm to mine the data to detect the association rules in each time period. We measured the similarity, difference and modified difference of the mined association rules based on three definitions and used the self-set rule matching threshold (RMT) to find all types of rules.
Figure 3 Different Types of Changes
5. System implementation
References
5.1. Data description
[1] Alex Sbesbunoff,”Winning CRM Strategies”, ABA Banking Journal,1999. [2] Vince Kellen, “CRM Measurement Frameworks”, Blue Wolf White Paper, p.4, 2002. [3] Ott, J., “Successfully development and Implementing Continuous relationship management,” eBusiness executive report, 2000, 26-30. [4] Adriaans P. and Zantinge D., “Data Mining,” Addson-Wesley, Harlow, 1996. [5] A. Berson, K. Thearling and S. Smith, “Building Data Mining Applications for CRM,” McGraw-Hill, 2000. [6] Jill Dyché, “The CRM Handbook: A Business Guide to Customer Relationship Management”, Addison Wesley Professional, 2002. [7] Swift, Ronald S., “Accelerating Customer Relationships Using CRM and Relationship Technologies,” Prentice Hall PTR,2001 [8] Song H.S., Kim J.K. and Kim S.H. “Mining the change of customer behavior in an internet shopping mall” Expert Systems with Applications 2001. [9] Shaw M.J., Subramaniam C., Tan G.W. and Welge M.E.,”Knowledge management and data mining for marketing”, Decision Support Systems 2001. [10] Grupe, G. H., Owrang, M. M., “Database Mining Discovering New Knowledge and Cooperative Advantage,” 12, 1995, 26-31.
The data for the empirical study are credit card customer data and spending records in 2003 and 2004 from banks. In light that the study focuses on the spending behavior of customers, all purchases of the same customer using more than one credit card were combined under said customer. There were altogether 10304500 and 1146200 pieces of purchase data for 2003 and 2004 respectively, which were matched using a program, and data of the same customer were kept. Finally, 1063000 pieces of data for each year were used as dataset for the study.
5.2. Weka data mining tool Weka (Waikato Environment for Knowledge Analysis) is a Java-based data mining tool developed by Waikato University. After loading the dataset into it, the preprocess function of Weka allows the user to input undesired attributes to prevent them from affecting the quality of extracted knowledge. Next, the user can use one of the three algorithms to mine the data: Classification, Clustering, and Association Rule.
6. Results In this section, we used Weka to run the model. First, we clustered the 2003 customers to identify the most influential group in term of business profit and find their behavioral pattern. The 2004 dataset was used to observe change of rules in different time periods. These steps can help decision makers
2
Proceedings of the 29th Annual International Computer Software and Applications Conference (COMPSAC’05) 0730-3157/05 $20.00 © 2005 IEEE
