現行營業稅自實施以來,營業人經常藉由漏開發票漏報營業收入、虛報進項稅額虛增費用等方式,藉此逃漏營業稅,稽徵機關如何以有限人力去防堵逃漏稅,特別是重大逃漏稅,為目前稽徵機關所需要努力的目標。 本文擬透過決策樹、區別分析、邏輯斯迴歸模式的建立,對加值型營業稅逃漏稅偵測,進行實證研究,其結果發現邏輯斯迴歸預測正確度89.3%,高於區別分析預測正確度87.7%,決策樹預測正確度更高達95.7%,希冀能夠使稅捐稽徵機關更有效辦理查緝以及防杜相關重大逃漏稅之功效,研析最佳之查核模式,供稅務相關機關參酌運用,藉由打擊不法,以遏止逃漏,維護租稅公平,增裕庫收。
From the practice of current business tax, the business entity usually evades the business tax against invoice and business income skipping and input tax and expense padding. How to block the tax skipping, especially for the severe tax skipping by limited human resource has been the biggest goal that the taxing authority shall strive for. This study detects the evasion of value-added tax and further carries out positive research against the creation of decision tree, discriminant analysis, logistic binary regression model. The result shows that the accuracy of logistic binary regression model is 89.3%, which is higher than the forecasting accuracy of discriminant analysis of 87.7%. The forecasting accuracy of decision tree is even up to 95.7%. We hope this can enable the taxing authority to investigate and prevent the relevant severe tax skipping more efficiently. The best investigation model can be applied by relevant taxing authorities to crack down on illegal practices and holding back tax skipping in order to maintain the taxation justice and increase the income.