利用資料探勘方法建構再分類信用評等模型

Using Data Mining Approaches to Construct a Reassigning Credit Scoring Model

黃思騰(Szu-Teng Huang)

指導教授 ： 林榮禾

國立臺北科技大學/管理學院/商業自動化與管理研究所/碩士(2007年)

摘要

隨著信用產業快速擴張與呆帳問題的產生，信用評等模型的建立成為重要的研究議題。過去幾年，統計以及人工智慧領域的學者紛紛建立許多信用評等模型來增加信用評等的準確率。本篇研究提出再分類信用評等模型(RCSM)，試圖解決信用評等與降低信用評等模型之型一誤差的難題。RCSM分為兩階段，第一階段建構以類神經網路為基礎的信用評等模型，用以將申請人分類為信用優良與信用不佳兩類別。第二階段建構以案例式推理為基礎的再分類信用評等模型，用以將被第一階段分類為信用不佳，卻有可能為誤判的申請人分類至有條件接受的類別，以降低型一誤差。案例分析採用UCI Repository of Machine Learning Databases的德國與澳洲信用資料，分析結果顯示，RCSM不僅在第一階段分類準確率高於其他四種比較方法，在第二階段也大幅降低RCSM的型一誤差，藉由型一誤差的降低，增加銀行的信用卡潛在客戶，並增加銀行收入與利潤。

關鍵字

資料探勘 ； 信用評等模型 ； 多變量迴歸 ； 類神經網路 ； 案例式推理 ； 型一誤差

並列摘要

Credit scoring model development became a very important issue as the credit industry has many competitions and bad debt problems. Therefore, most credit scoring models have been widely studied in the areas of statistics and machine learning to improve the accuracy of credit scoring models during the past few years. In order to solve the classification problems and decrease the Type I error of credit scoring model, this study presents a reassigning credit scoring model (RCSM) involving two stages. The first stage is to construct an ANN-based credit scoring model, which classifies applicants with accepted (good) or rejected (bad) credits. The second one is trying to reduce the Type I error by reassigning the rejected good credit applicants to the conditional accepted class by using the CBR-based classification technique. To demonstrate the effectiveness of proposed model, RCSM is performed on two credit card dataset obtained from UCI Repository of Machine Learning Databases. As the results indicated, RCSM not only provides more accurate credit scoring than other four common used approaches, but also increase business revenue and decrease the Type I error of scoring system.

並列關鍵字

Data mining ； Credit scoring model ； MARS ； ANNs ； CBR ； Type I error

參考文獻

[2] Durand, D. (1941). Risk elements in consumer installment financing, National Bureau of Economic Research, New York.

[3] Freeman, J. A., & Skapura, D. M. (1992). Neural networks algorithm, application and programming techniques. MI, USA: Addison-Wesley.

[5] Hecht-Nielsen, R. (1990). Neurocomputing. Menlo Park, CA: Addison-Wesley.

[7] Bardos, M. (1998). Detecting the risk of company failure at the banque de france. Journal of Banking & Finance, 22, 1405-1419.

[8] Barney, D. K., Graves, O. F., & Johnson, J. D. (1999). The farmers home administration and farm debt failure prediction. Journal of Accounting and Public Policy, 18, 99-139.