標題: | 運用資料探勘方法進行信用卡靜止戶預測 Using Data Mining Techniques to Predict Credit Card Dormant Account |
作者: | 黎益忠 Li, Yi-Chung 劉敦仁 Liu, Duen-Ren 管理學院資訊管理學程 |
關鍵字: | 資料探勘;信用卡;靜止戶;支援向量機;隨機森林;Data Mining;Credit Card;Dormant Account;Support Vector machine;Random Forests |
公開日期: | 2014 |
摘要: | 台灣信用卡市場競爭日益激烈,因此留住高忠誠度及高貢獻度客戶已成為每個銀行重要的課題。根據銀行局四月份新聞稿指出,截至103年4月底為止,共計有36家信用卡發卡機構,總流通卡數約3,646萬張,亦即全台每位居民持有1張以上信用卡。
本研究將針對國內某知名銀行在2011年1月至2012年12月期間,連續24個月都有消費,且每個月的消費金額都達一萬元以上,定義為高忠誠及高貢獻客戶做為研究對象。以上述客戶來觀察在2013年1月至2014年2月期間,連續兩個月沒消費的客戶,即定義為靜止戶,我們希望透過資料探勘方法,來找出相關客戶的特徵。
本研究以消費金額及筆數來發展出可測得之指標,再運用支援向量機以及隨機森林兩種分類預測方法,來找出靜止戶的特徵,希望可以在客戶轉變為靜止戶之前,將相關訊息提供給企劃人員,預先對客戶進行促動來避免客戶流失,持續讓客戶使用銀行信用卡,維持客戶消費動能,來增加客戶與銀行之間的關係以及黏著度,進而可對客戶進行跨售,以增加銀行從客戶身上持續獲利之可能。 It is a hypercompetitive environment in Taiwan credit card market, therefore, retaining high loyalty and contribution card holders has become one of the most important issues for each bank. According to the press release from Taiwan Banking Bureau on April 2014, there are 36 issue banks and thirty-six million credit cards in circulation as well as each resident holding more than one credit card in Taiwan. The aim of this research is to discover the pattern of high loyalty and contribution customers, who used credit card to spend more than 10 thousand NTD each month from 2011 to 2012, transformed to dormant customers, who had no transaction for two consecutive months from January 2013 to February 2014, by applying data mining techniques. In this research, Support Vector machine and Random Forests as well as the indicators developed from transaction amounts and frequencies have been employed to distinguish the characteristics of dormant customers. The information would be provided to product manager who could launch retention campaigns to maintain better relationship with card holders and reduce the probability of being dormant accounts. Furthermore, it could increase the possibility of enlarging the profit of the bank. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#GT070163422 http://hdl.handle.net/11536/76019 |
顯示於類別: | 畢業論文 |