本研究以還款績效的觀點，分析小型信用貸款中申貸者的特性，讓銀行放款的依據除了判斷正常戶與否之外，進一步以還款績效與風險區分出不同群組的申貸者，以期作不同的放款策略；同時將個人基本變數 、該銀行內徵信資料以及聯合徵信資料變數 作統計性分類，篩選出代表性因素，研究這些因素如何影響各還款績效群組。 研究發現，申貸者可以區分為「還款能力平穩—逾期風險低」、「還款能力優良—逾期風險中」、「還款能力低下—逾期風險高」這三群。而從影響各群組的因素中可以瞭解，「還款能力平穩—逾期風險低」群組，多為各方面信用持平良好的申貸者；「還款能力優良—逾期風險中」群組，多為具有理財管理特質、財務狀況良好的申貸者；「還款能力低下—逾期風險高」群組，多為具有債務因素、信用表現不佳、申貸動作頻繁的申貸者。 在放款利潤與風險方面，對三個群組之申貸戶分別採用不同方法放款，可以作到讓銀行對較少申貸戶放款並且可提升利潤並且改善損失。進行多元羅吉斯迴歸模型分析可以發掘出具影響力的因素，針對這些因素來進行分群後並採差異化放款方法，也可以作到對較少申貸戶放款並且能提升利潤以及降低損失的效果。由於因素代表具解釋性變數的歸納，配合這些具預測機能的因素及變數分群訂定差異化授信政策，有助於防範風險於未然。 / This research analyses the characteristics of small-scale credit loan applicants on the persepective of repay performances，allowing the banks not only to discriminate between good and bad applicants but also to establish different lending tatics for applicants of different repay performance groups。We also analyse the personal characteristics and joint credit informantion of these applicants to sieve out the representative factors，and study how these factors affect the repay performance groups。 Our research discovers that the applicants can be discriminanted into three groups：「low but steady repay ability—low overdue loss」、「good repay ability— acceptable overdue loss」、「very low repay ability—high overdue loss」。We can learn from those factors，that most applicants grouped as 「low but steady repay ability— low overdue loss」also have good credit qualities in other aspect；applicants grouped as 「good repay ability—acceptable overdue loss」 have finance management concept and good financial condition；applicants grouped as 「very low repay ability—high overdue loss」have debt burdens and bad credit qualities。 As for the revenues and riks，we can improve the profit and loss with fewer applicants by taking differenct lending policies to those three groups。By using multinomial logistic regression，we can discover those factors who has significant effects and use these factors to cluster applicants into groups and to adopt different lending policies for these groups。Because those factors represent the induction of the variables which can explain the applicants’ behaviors，we can somehow prevent the risks by establishing different policies with the coordination of these factors and clusters。