Automated Round-trip Detection Based on Data Mining Methodologies

• Main Authors: Wen-Hsiung Wu, 吳文雄
• Other Authors: Li-Chen Cheng
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/56263052743386293456

碩士 === 東吳大學 === 資訊管理學系 === 99 === Tax revenue is the major source of government income. However, as information technology is getting advanced, enterprises keep trying to find new methods to evade tax. Current business taxation in Taiwan can be classified to Value-Added and Non-Value-Added types. The above two types of taxation adopt multi-stage sales tax processes. The former levies the tax based on the amount of added value and can draw up the uniform invoice as the certificate to deduct the drawback. The fictitious company is a major source of business tax evasion. However, the patterns of fictitious companies to evasive tax are varied and complex. The round-trip is the most serious one which has badly affected government's tax revenue. In recent years, there are several corporate fraud cases happened. The round-trip is the most frequent manipulative scheme. It always depends on the experiences of tax inspectors to detect fictitious criminal companies. When inspectors use tax information system to check on the correction of all the detailed purchase-and-sale items of business declaration, most of them could be overwhelmed by the huge amounts of abnormal outputs. It is inefficient and a waste of inspecting labor. The present advanced procedure delivers the possible multi-cycle transaction detailed list by the computer program. However, there are too many suspicious items to find out all of them. Therefore, a well inspection priority for the items will make the efficient inspection and reduce the revenue losses. In this study, we use the income/outgoing tax of business's transactions and characteristics of abnormal fictitious company data to find out suspicious business transactions. The social network analysis is also used to review the behavior of transactions. Moreover, classification algorithms and data mining technology are adopted to identify inspection priorities that will help inspectors to do a full check of the suspected companies. The proposed method in this study can check out many levels of round-trip transaction from criminal fictitious companies, so that the tax auditors can find out the illegal fictitious company to keep business taxation fair. Business tax data, collected from National Tax Administration of Northern Taiwan Province, was used for data mining process.