傳統銀行業授信處理，多位學者採用統計模式如 MDA,LOGIT,PROBIT,REGRESS ON等，卻產生一些應用上不恰當的困擾。本文試 圖利用多準則決策工具一AHP，建置一個金融授信決策模式，以求改善這類統計 模式的諸多缺失。為了讓決策模式更加完整，文中方修改傳統5P的授信因素，使 模式內之決策因素具有較高的獨立性。 此外，為了使泱策者在運用此一模式時，能充分掌握資訊來源，建議在現有金融 自動化之基礎上建立一個授信決策資訊通道架構。經由此一架構，使用者將可自 銀行內部系統內擷取授信戶日常交易之彙整資料，及外部金融週邊單位相關之信 用資料。 由於考量了決策因素的獨立性，子系統將更容易單獨建置;針對財務分析子系 統，本文採用了Miller[19]的專家知識為籃本，在個人電腦上建置了一個專家系統 的雛型。

　　　　　Traditionally, the researchers used a few statistical models like MDA, LOGIT, PROBIT, REGRESSION, ...etc, to build the commercial loan decision models. However, they got some criticisms from practical viewpoint. Trying to amend these drawbacks, we use a multiple-criteria decision making tool-AHP (Analytic Hierarchy Process) to weave a commercial loan decision model. In order to improve the dependence of each decision factor, we also modify the popular 5P Principle in our model. Besides, in order to support more accurate information to the model, we suggest an information channel architecture building on the infrastructure of the local banking automatic systems. Under such circumstances, the user not only can get the summary account status of the applicant from the inner system, but also can retrieve some credit information from outside financial institutes. Finally, adopting the cause-and-effect ratio approach from Miller [19] as the basic rule base, we implement a prototype Expert System called AFRESS on a Personal Computer.