房屋貸款是國內各銀行(包括本國一般銀行、外國銀行在台分行及中小企業銀行)最主要的消費性貸款業務及獲利來源,所以房屋貸款的逾放比重的大小,影響銀行收益甚巨。 隨著逐漸開放的金融政策,金融體系邁向自由化與全球化。不論是公、民營銀行,為爭取市場占有率,會使銀行授信原則有所鬆動,導致授信品質下降,造成金融機構逾期放款比率呈逐年增加之趨勢。 本研究嘗試以決策樹模式、貝氏分類法及粗糙集模式作為輔助工具,來探討預測值,並找出更具代表性的決策模式。期望能將研究結果建立一套公平和客觀之消費者房屋貸款的預警評估模式,藉此提供銀行未來針對貸款之個人調整授信政策之參考依據,以增加違約戶篩選的正確性和減少銀行的損失。 本研究的實證結果顯示,決策樹模式有最高的整體正確率,達到85.3%。
The mortgage loan is the highest proportion of various consumptive loan services for Taiwan financial institutions, includes domestic banks, the local branches of foreign banks and medium business banks. The revenues of the Taiwan financial institutions is significantly related to the proportion of overdue mortgages. With gradual opening of financial policy, financial system heads for liberalization and globalization. Local banks, private or public, intend to increase market share so that the crediting principles are somewhat loose and quality lowers as well. As a result, the bad debt rate is on the increase year by year. This study attempts to Classification tree model, Bayesian classification and the rough set model as a supplementary tool to explore the predictive value, and to identify more representative decision-making model. The purpose of this study is to build a fair and objective default prediction models for mortgage loan. The findings of this study can provide useful information to banks for making decision about credit policy of personal mortgage in the future. Besides, it can increase the correctness from sieve of default customer and decrease the loss of banks. The empirical results show that Classification tree model has the highest overall accuracy rate reached 85.3%.