運用資料挖掘技術於信用卡顧客關係管理之研究

Applying Data Mining Technology to Enhance Customer Relationship Management in Credit Card Business.

王景聰(Jing-Chung Wang.)

指導教授 ： 蔡介元博士

元智大學/工程學院/工業工程與管理學系/碩士(2003年)

摘要

隨著金融控股公司如雨後春筍般的成立，集團式的金融經營模式讓金融服務更為多元化，一次購足(One Stop Shopping)式的金融服務更是業者搶佔商機以獲得消費者喜愛的關鍵，也因此引發更全面性的金融激戰。在眾多消費性金融產品中，信用卡則是受到消費者的青睞，大幅的取代現金的使用，致使銀行爭相投入信用卡發卡市場；也因此，以顧客為導向的「顧客關係管理」便繼網際網路風潮後，成為全世界企業主所共同關切的熱門話題。
發卡機構為做好顧客關係管理，目前的研究大都集中於信用卡核卡準則分析及偽卡交易的偵測，或將現有的信用卡資料作一個分類或輪廓分析。然而，對於信用卡獲利有實質貢獻的族群探討，及此一族群因時間變動的消費行為模式，這兩個重要議題卻是一知半解且被乎略，以致於管理者做決策時產生困擾。有鑑於此，本研究針對信用卡VIP族群進行二個不同時間區段的消費行為分析與探討。首先，我們將所要分析的二個不同時間區段的VIP資料（包含顧客基本資料及消費資料）篩選出，接著利用關聯法則演算法進行資料的挖掘，以瞭解每一時間的關聯規則。接著我們將挖掘出的二個不同時間區段關聯規則，根據三種不同類型的定義（即Emerging Patten Rule、Unexpected Change Rule、Added／Perished Rule）計算其相似性、差異性及調整差異性的量測值，同時運用自訂的規則比對門檻（RMT）的值以找出所有不同類型的規則。之後，再運用每種類型的程度改變之量測值找出顯著改變的規則。
在本研究中，我們運用COBOL、JCL、IBM Intelligent Miner工具、VB程式、及Access資料庫成功地將上述的系統開發完成。在我們的實驗及分析過程中，我們發現大多數信用卡VIP族群的消費多集中在百貨公司、超級市場、加油站、餐飲業、家庭用品及休閒娛樂。此外，本研究所找出的顯著Emerging Patten Rule、Unexpected Change Rule、Added／Perished Rule中，根據此三種類型規則的趨勢變化，我們可以有效地偵測出顧客的目前行為以及可能改變的行為趨勢，讓管理者能在大量資料中發現一些可能的改變，及早提供顧客所需的產品及服務以擴大顧客群及避免顧客的流失。同時在瞭解顧客的可能改變及消費趨勢走向後，決策人員便能將有限的資源做最有效的運用，以制訂出適合不同顧客群的行銷策略。

關鍵字

資料挖掘 ； 關聯法則 ； 信用卡 ； 顧客關係管理 ； 決策管理

並列摘要

In the recent year, financial groups in Taiwan are allowed to provide diverse financial services. One stop shopping service has become the key on which a financial enterprise can survive under extensive competitions. Among all consumer-banking products, credit card is most attractive to consumers. The magnificent increase in credit card market leads banks to put more efforts in understanding the behavior of customers. Therefore, customer-oriented Customer Relationship Management (CRM) turns into one of the most important topic for the enterprises.
In order to conduct better CRM, most credit card related researches focus on issuing principle decision, fraud transaction detection, and customer profile description. However, few efforts are conducted upon the discussion of the consumption behavior changes over time of the consumer group, which contribute most in the credit card business. To bridge the gap, this research aims at the analysis of the consumption behaviors of the very important person (VIP) in two different time intervals. First, VIP data, including customer profiles and purchase transactions, of two time intervals are screened for this analysis. Then apriori algorithm is used to generate association rules for each time interval. The rules in the two time intervals are then calculated for their respective similarity measure, difference and modified difference measures according to three types of rules : Emerging Pattern Rule, Unexpected Change Rule, Added/Perished Rule. Meanwhile, the rules for all different types are found using a user-defined rule match threshold (RMT). Finally, significantly changed rule set is found using evaluation for degree of change measure for each types.
The system mentioned above has been successfully built using COBOL, JCL, IBM Intelligent Miner, VB program, and Access database. In this study, it was found that the VIP group often visits department stores, supermarkets, gas stations, restaurants, home appliance stores, and recreation sites. In addition, the customers’ consuming behaviors are detected based on the three mentioned rule types. This allows managers to find critical changes of VIP customers from a large amount of data. The decision planners can benefit by the finding and maximize the utility of their limited resources to establish suitable marketing strategies for different types of customers.

並列關鍵字

Data Mining ； Association Rule ； Credit Card ； Customer Relationship Management ； Decision Management

參考文獻

[15]張文彬（民90）。顧客關係管理的核心活動在企業界應用過程之探討。私立中原大學企業管理學系研究所碩士論文。

[3]Areawal, R. Srikant, R.（1994）. Fast Alorithm for Mining Association Rules in Large Databases, In Proc Int’l Conf. VLDB, pages 487 — 499 , Santiago, Chile.

[6]Chan, P.K.; Fan, W.; Prodromidis, A.L.; Stolfo, S.J.（1999）. Distributed data mining in credit card fraud detection. IEEE Intelligent Systems, 14（6）, pages 67-74.

[7]Cohen, E , Datar, M. Fujiwara, S, Gionis, A, Indyk, P, Motwani, R, Ullman, J, D, Yang, C.,（2000） Finding Interesting Associations without Support Pruning, Data Engineering, Proceedings. 16th International Conference on Pages 489 — 500.

[10]Fayyad, U.M.,Shapi, G.P., Smyth,P. and Uthursamy, R.（1996） Advances in Knowledge Discovery and Data Mining. AAAI Press/The MIT Press, Menlo Park, CA.