Fictitious Company Detection Based on FKMS Framework

• Main Authors: Yu-Mei Yang, 楊玉美
• Other Authors: Yi-Chuan Lu
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/77167411041805755420

碩士 === 元智大學 === 資訊管理學系 === 98 === The investigation of a newly company tax registry is the first and most important step for preventing a fictitious company form being registered. This research is developed on the FKMS platform with DMAIC methodology which provides a process for knowledge workers to Define, Measure, Analyze, Improve and Control in their activities. The business tax data of business people, who are registered in wholesale and retail trade from 2003 to 2008, are used in this study. By using the back-propagation neural network technique, an offensive system is established to distinguish those fictitious companies from other normal companies. The network topology consists of a single hidden layer with 9 processing units. The learning rate is set to 0.4 and the accuracy rate is up to 11.3132%. Compared with the current 2.38% accuracy rate, it is greatly enhanced and improved. From the sensitivity analysis, the results shows that the (1) business owners (1) with a large amount of taxes owed, (2) as other company’s owner at the same time, and (3) were used to be in charge of other abnormal business, would be the important factors for case selections.