Application of data mining techniques for VAT-Registered Business compliance

Abstract

World Bank recommends that Indonesia lower the turnover threshold required to be a VAT-Registered Business from Rp. 4,8 billion to Rp. 600 million to increase VAT-Registered Businesses numbers which will also increase VAT revenue. The number of VAT-Registered Businesses will be significantly increased, which will push Directorate General of Taxes to determine the correct audit priority because it is impossible to audit all taxpayers. This study aims to form a prediction model for formal compliance of VAT-Registered Businesses in the Sampit Tax Office towards 1270 VAT-registered Businesses as of December 31, 2019, which are classified as low-risk VAT-Registered Businesses. The prediction model will be useful for determining audit priorities for certain taxpayers. This study uses a qualitative method using the RapidMiner application and decision tree technique in making prediction models for VAT-Registered Business compliance. The model made has Prediction Efficiency of 67,9%, reduction in Examination Effort by 63.67%, and Strike Rate of 85.99%. The model made is used to predict new VAT-Registered Business data which registered in 2020 and predicts 76 VAT-Registered Businesses will be compliant and 7 VAT-Registered Businesses will not be compliant