本文目的在建構一區別虛設行號營業人之預測模式,鑑於此,研究資料以財政部臺灣省中區國稅局(以下簡稱中區國稅局)於97年6月聯合臺灣彰化地方法院檢察署、彰化縣警察局員林分局共同破獲之饒玉鱗虛設行號集團,轄屬中區國稅局之營業人為樣本,藉由觀察該等營業人每期營業稅申報資料,運用Logit迴歸分析、區別分析找出判別虛設行號之特徵因子,分別建立預測模式。 實證結果顯示,營業人申報加值率、營業稅申報期數、取得進項發票張數、平均每張進項發票金額等變項,為判別營業人是否為虛設行號之顯著因素,並建立了兩分類模式,實證顯示,邏輯斯分析、區別分析的預判率分別達到100%、99%,此模式有助於稽徵機關及早掌握虛設行號營業人不法行為,節省稽徵人力,有效達到遏止逃漏之目的。
The purpose of this study is to construct the prediction model for discriminating the fictitious company. In view of this, the sample data was selected from the fictitious companies in the National Tax Administration of Central Taiwan Province, Ministry of Finance (hereinafter referred to as Central National Tax Bureau) that Central National Tax Bureau jointed Changhua District Prosecutors Office and Yuanlin Branch of Changhua County Police Headquarter to uncover Rao Yu-Lin Fictitious Company group commonly on June, 2008. According to the observation of fictitious companies’ each transaction tax(business profit tax or sales tax) declaration data, and use logistic regression model and discriminant analysis model to find out the characteristic factors of fictitious company judgment for constructing predictive models each other. As the empirical results, some variables of fictitious company would be the significant factors to judge whether the business person is fictitious company or not, such as added-value tax declaration, business tax declaration terms, income invoice quantities acquirement and each income invoice amount average, and then construct two kinds classification models. The empirical results showed that the correct classify rate of logistic Regression model and discriminant analysis model are reached to 100% and 99%. These models will help National Tax Administration to monitor fictitious companies’ illegalities early and save audit manpower to accomplish the purpose of tax evasion preventive effectively.