本文旨在利用層級分析法(簡稱AHP)、區別分析與logit模式，探討銀行授信客戶 的信用評等。 實證結果發現，就 AHP 方法而言、正常授信案件在四種實驗組別之平均區別 能力達 82%，但違約案件的平均區別能力則僅達 70%；從區別分析法觀之，區別模式對正常 客戶的平均區別能力為 85%，對違約客戶的平均區別能力略低，僅達 75%； 由 logit 模式 得知，該模式對正常客戶的平均區別能力為 91%，對違約客戶的平均區別能力為 85%。故整 體而言，logit 模式優於區別分析，而區別模式又優於 AHP 方法。

　　　　　By use of AHP, discriminant analysis and logit model, this study emphasizes on credit rating of bank loan client. Based on the empirical evidences in AHP approach, the paper found that the mean correcting rate for regular customers is 82%, higher than that of the default customers, in which the correcting rate is about 70%. In view of discriminant analysis, the mean correcting rate for regular clients is 85%, above on 75% of the correcting rate for the default clients. From the empirical test for logit model, the mean correcting rate of regular and default clients are 91% and 85%. On the whole, the research found that the logit model is better than discriminant analysis, and the discriminant approach is superior to AHP method.