Discriminant analysis for Military Unit Procurement Fraud Prosecution Types
| Autor: | Chung-Wen Hao, 郝崇文 |
|---|---|
| Rok vydání: | 2009 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 98 The primary work to prevent procurement corruption is to locate the loopholes that cause corruption. Therefore, various institutes of the government resort to the experiences from experts and case studies in procurement frauds in order to analyze and find out patterns of behavioral misconducts in procurement. However, it needs effective classification methods to detect the prosecution types so that prevention and reaction measures can be taken. Thus, this research aimed at find out a better way to detect the prosecution types. Firstly, this research issued open questionnaires among all inspector officers of R.O.C. Armed Forces, and retrieved feedbacks on causes of procurement frauds in military unit, and gathers procurement fraud cases since 2006. The research samples are 178 procurement fraud cases enlisted in military courts and prosecutor offices as well as in Procurement Auditing Office in Ministry of Defense. The samples are divided into two groups for analyses: 90 cases of training group and 88 cases of experiment group. Secondly, the classification rules of procurement fraud prosecution types are developed with three discrimination methods, i.e., ‘standardizing discriminant analysis after factor analysis’, ‘advanced discriminant analysis after factor analysis’, and ‘advanced discriminant analysis’ using training data. Thirdly, classification and correct detected rate is processed using the classification rules developed. The results showed that correct detected rate are higher than 90% among three methods and advanced discriminant analysis come out the best correct detected rate among the three methods. Finally, based on discriminant rules set up by the three discrimination methods using training group data, the research classified the experiment group data, which results in correct detected rate are 87.50%、88.64%、92.05% for three methods respectively. The advanced discriminant analysis come out the best correct detected rate among the three methods. Keywords: procurement fraud, factor analysis, discriminant analysis, advanced discriminant analysis |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|