本篇旨在檢視與探討S銀行目前使用之房貸評分表該使用之基本屬性子模型相關變數因子，於經過這些年在房貸市場快速成長與景氣復甦下是否須予以調整、新增？同時對於現行房貸業務與銀行政策進行探討，以期待於未來實務運用上給予較客觀之統計分析參考。 本研究採用Logistic迴歸分析，包含卡方檢定、t檢定與F檢定及對各因子（組合）之樣本平均違約率與中位數進行統計分析與比較；另同時進行兩實證模型之PD變異數相同檢定與PD平均值相同檢定。 首先針對基本模型中現行使用之變數因子該＂分類與配分＂進行初步檢視與了解；再透過衍生模型1針對現有之因子進行分類與配分上之調整，已達到適合目前市場業務與政策上之規範與見解。 並藉由衍生模型2以新增（組合）因子之角度，解決原變數因子該較薄弱之解釋力，同時配合現行業務之規範，以期達到模型精進之成效。 本研究的第二個重點擬分別從擔保品所屬縣市、借款金額佔淨值比率（LTV）、每坪單價、建物面積（含公設）等不同面向與組合進行實證統計與檢定；最終配合預測之PD值以BCG Matrix概念對現行業務進行分析，並期許於未來業務發展與政策修訂時給予適切與客觀建議。 Given the rapid growth of the mortgage market and the economic recovery these years, the research reexamines variable factors in the basic submodel of Bank S’s current mortgage credit scorecard. We discuss if they should be adjusted and if new variables should be added, and we also probe into its current mortgage services and lending policy, hoping to offer objective statistical analysis that would shed light on future practices. First of all, we conducted a preliminary examination of “the categorization and the ratio” of variable factors in the basic model before they were further adjusted through Derived Model 1 to better meet current services and policy. Low explanatory power of the original factors was then tackled in Derived Model 2 by adding (combining) factors. Regulations of the current services were also addressed to improve the model. Secondly, we conducted an empirical statistical test on factors and combination of factors, such as locations of the secured properties, their loan-to-value ratio (LTV ratio), price per ping (a ping is an area measure equal to 3.3057 square meters), property size (including public facilities). Finally, with the estimated probabilities of default (PDs) in the BCG Matrix, current mortgage services were analyzed with an aim to provide suitable and objective suggestions for future service development and policy revision.
