金管會為減輕卡債族的還款負擔,於2010年4月26日規定銀行對持卡人連續使用循環信用達一年以上,最近一年內於聯徵中心無帳款遲繳或信用不良紀錄,提供個人信用貸款及信用卡分期二種還款方案。由於金管會對信用卡的措施,銀行須同時考量收益與風險管理的問題,依據申請人信用狀況,審慎核卡,並給予合適的信用卡額度。因此,發卡銀行須建構信用評分機制,以降低信用卡呆帳率。 本研究目的為探討那些自變數具有模型解釋能力,並建立信用卡預測模型,測試婉拒戶有多少比率可成為授信戶,及聯徵外部分數與銀行往來內部分數,是否具有模型解釋能力,使A銀行獲得商機。 研究對象為A銀行的信用卡客戶共49,331筆,其中包含已核卡樣本及婉拒樣本,資料期間自2008年5月至2009年8月,採用聯徵中心(JCIC)提供給國內A銀行之信用資料,及A銀行在進件系統中的客戶基本資料及與銀行往來資料共24項自變數,做為評估分析的依據,透過羅吉斯迴歸分析法,探討影響信用卡審核的重要因素為何。本研究結果彙整如下: 一、經實證結果,共萃取出19項自變數具有模型解釋能力,並建立 信用卡預測模型。 二、運用本研究建立之模型,測試婉拒戶有94.92%可成為授信戶, 婉拒比率將降低,使個案銀行獲得商機。 三、聯徵外部分數以及聯徵外部分數與銀行往來內部分數兩者合計之 分數具有模型解釋能力。
In order to reduce the burden of card revolvers, Financial Supervisory Commission regulates the banks to provide either personal credit loans with amortized payment schedule or credit loans to credit balance bearers with more than one year on-time payment history. Accordingly, banks need to consider both the revenue and risk management based on the applicant’s credit worthiness before approving the credit line. Thus, card issuing banks should build up the scorecard system to reduce the delinquent rate. The objective of this research is to study the explanatory power of the variables in the credit model, and build up the credit card’s predicting model. To testify how much ratio the rejecting applicants to explain the model, therefore, the bank A could make profit. A sample of 49,331 credit card applicants, including approved and rejected, were collected from Bank A. Sampling period started from May 2008 to August 2009. A set of 24 explanatory variables were developed based on external data from JCIC and internal credit data of Bank A. Using the logistic regression, we explore the important factors of credit card characteristics. The results summarize as follows: 1. There are 19 independent variables that have the ability to explain the model, and to build up the predicting model of credit card. 2. According to the model, there are 94.92% rejected applicants who could become the approved. Thus, the bank could make profit if the rejecting rate decreases. 3. The sum of JCIC outer score and inner score has the ability to explain the model.