本研究除了傳統的銀行授信基本變數、法拍屋特徵變數外，進一步將投資客特徵相關變數，加入模型來建構邏輯斯迴歸模型，可以有效提高整體模型準確率。其中有關投資客特徵變數，「申請人現居地址與投標不動產標的，二者是否有地緣關係」、「借戶最近3個月被其他銀行因新業務查詢的次數總計高達三次以上者」、「借戶於申貸時已有一筆以上擔保貸款授信餘額資料」、「借戶於申貸時依3年內購置不動產結案資訊高達三筆已者」等四項變數呈顯著水準。也就是當納入投資客特徵相關變數之後，整體模型預測力由80.5%提高到91.0%；雖然正常戶之預測準確率由97.8%稍稍降低為95.4%，但是房貸逾期戶之預測準確率卻由11.5%大幅提高為73.5%。說明銀行在建構法拍屋違約信用模型時，由於法拍屋市場充斥許多投資客，銀行更應將此投資客特徵相關變數納入，可以有效預測法拍屋房屋貸款逾期戶的機率，除了可以降低銀行的信用風險，亦可符合管理當局要求的資本規定。

This study extends the Foreclosed House credit scoring system with further five Speculator characteristics, i.e. ＂debt ratio＂, ＂whether the loan's residence and tender object is in different locations＂, ＂the borrower of last three months by other banks because of the number of new mortgage query totaled more than three times＂, ＂the borrower to apply for loans has more than one loan credit ＂and ＂the borrower to apply for loans had paid off the mortgages totaled more than three times within three years＂, in the logistic regression model to build a dichotomous prediction credit scoring model which can be adopted by financial institutes to prevent default risk of Foreclosed House mortgage loan and improve the quality of risky asset. The empirical results indicate that the overall accurate predict rate of the model with five Speculator characteristics (80.5%) is higher than the model without five Speculator characteristics (91.0%) indicating that characteristics of speculator has the most ability to predict default rate.