Evaluating credit scoring models
by Vesarach Aumeboonsuke
Title: | Evaluating credit scoring models |
Author(s): | Vesarach Aumeboonsuke |
Advisor: | Dryver, Arthur L, advisor |
Degree name: | Doctor of Philosophy |
Degree level: | Doctoral |
Degree discipline: | Finance |
Degree department: | School of Business Administration |
Degree grantor: | National Institute of Development Administration |
Issued date: | 2011 |
Digital Object Identifier (DOI): | 10.14457/NIDA.the.2011.22 |
Publisher: | National Institute of Development Administration |
Abstract: |
Evaluating the credit worthiness of credit seekers is a crucial process for financial institutions simply because their existence largely depends on how such a process is conducted. Financial institutions use a variety of credit scoring methods and a variety of criteria to select the best credit scoring methods. The primary purpose of this research is to evaluate the performance of some of the existing popular credit scoring methods that are widely used by financial institutions. The credit scoring methods to be considered for comparison purpose include logistic regression, discriminant analysis, and recursive partitioning. Several statistical criteria to be considered for evaluation include the Kolmogorov-Smirnov statistic (K-S), the Gini coefficient, and odds ratio at various cut-offs. Much research in the past has compared credit scoring methods through using sets of real-world data. In this paper, however, the comparison of the credit scoring methods has been done by using a set of data generated through simulation in order to acquire extensively representative and sufficiently effective samples; in this way, it is possible to compare and validate the performance of the classification models on the population. This paper simulates the data sets of the population, draw samples from each population set, runs each credit scoring method on each data set, computes the K-S, Gini, and odds ratio for each model for each data set, compares the cross-validation with the K-S, Gini, and odds ratio at various cut-off points, and evaluates the performance of different credit scoring models across different methods, across samples with different ratios of “goods” to “bads”, and across samples drawn from the population with different characteristics. The findings of this research will be useful for financial institutions, especially commercial banks, because they present evidence of how well each credit scoring method can predict the credit score of loan applicants. Banks make lending decisions based on such credit scoring systems, and the lending decision is crucial because it is the source of their revenue. If the bank accepts the applicant that is going into default, then it will have a bad loan, which results in loan losses. On the other hand, if the bank rejects the applicant that is not going into default, then the bank has lost the opportunity to gain more revenue from that applicant. Therefore, ideally, banks would like to use a credit scoring model that has a stable and reliable predictive power across different characteristics of populations and sample sets. |
Description: |
Thesis (Ph.D. (Finance))--National Institute of Development Administration, 2011 |
Subject(s): | Credit scoring systems |
Resource type: | Dissertation |
Extent: | ix, 122 leaves : ill. ; 30 cm. |
Type: | Text |
File type: | application/pdf |
Language: | eng |
Rights: | This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. |
URI: | http://repository.nida.ac.th/handle/662723737/574 |
Files in this item (CONTENT) |
|
View ทรัพยากรสารสนเทศทั้งหมดในคลังปัญญา ใช้เพื่อประโยชน์ทางการเรียนการสอนและการค้นคว้าเท่านั้น และต้องมีการอ้างอิงแหล่งที่มาทุกครั้งที่นำไปใช้ ห้ามดัดแปลงเนื้อหา และทำสำเนาต่อ รวมถึงไม่ให้อนุญาตนำไปใช้ประโยชน์เพื่อการค้า ไม่ว่ากรณีใด ๆ ทั้งสิ้น
|
This item appears in the following Collection(s) |
|
|