Using Machine Learning to Identify Incorrect Value-Added Tax Reports

Authors
G. Van Bree, S. Staudinger, F. Burgstaller, F. Schiff, C. Schütz
Paper
Stau24a (2024)
Citation
Proceedings of the 30th Americas Conference on Information Systems (AMCIS 2024), Salt Lake City, Utah, August 15-17, 2024, Association for Information Systems (AIS), 2024.
Resources
Copy  (In order to obtain the copy please send an email with subject  Stau24a  to dke.win@jku.at)

Abstract (English)

Many companies and organizations worldwide have the legal obligation to periodically file value-added tax (VAT) reports. Typically, VAT reporting is done manually by accountants. The accountant indicates the amount of tax that has to be paid by assigning a tax code to a respective accounting document. Incorrectly assigned tax codes may lead to an incorrect amount of VAT that is reported to the authorities. Either the company is paying more VAT than necessary, which is at the company’s expense, or less VAT than necessary, which violates the law and may result in additional fines. We propose a system that uses machine learning to identify incorrectly assigned tax codes for accounting documents in order to help companies stay compliant with current tax law. Our system was evaluated on a real-world case of an internationally operating manufacturing company from Austria, which included data on over 70 000 invoices.