消費者信用貸款違約風險評估模型之研究－－以CART分類與迴歸樹建模

On the Research of Default Risk Model of Consumer Credit Loan-Using CART (Classification and Regression Tree)

本研究旨在找出影響消費者信用貸款違約之重要變數，並藉以建立消費者信用貸款違約風險評估模型，以期做為銀行在信用貸款風險管理上之參考依據。本研究以相關分析、因素分析及主成份分析來選擇及整合對信用貸款違約有重要影響力之變數，並以分類與迴歸決策樹（CART）進行預測違約模型之建構，最後經由模型預測能力指標、ROC曲線與Cumulative Lift曲線之綜合比較後，選取出消費者信用貸款違約風險預測之最適模型～「違約逾期相關程度法－決策樹」模型。研究結果顯示，此模型整體的預測正確率可達77%以上。

research is for the purpose of discovering the important variables which will affect the consumer credit loan. So as to and establishes the default risk model of consumer credit loan. This research uses the correlation analysis, factor analysis and principle component analysis to integrate those important effective variables. And the CART (Classification and Regression Tree) method is adopted to build up a default risk model of consumer credit loan. Three kinds of variable selection models including "correlated with default", "factor analysis", and "principle analysis" are proposed and compared. By comparing the Accuracy rate, ROC Curve, and Cumulative Lift Curve of those models, the "correlated with default" model is chosen to be the adaptable model. The Accuracy rate of this chosen model is high than 77 percent.