Most public companies are required by law to have their financial statements audited annually by external auditors. The purpose of an annual audit is to strengthen the confidence of prospective users in the financial statements. The auditor has to express an opinion on whether the financial statements have been prepared in accordance with applicable accounting standards. The risk-oriented audit approach is designed to cope with the ever-increasing complexity of the companies to be audited.
Transactions that do not originate from well-controlled processes. Such unusual business transactions may involve an inherent risk of material misstatements. Audit standards require an analysis of accounting data on a detailed level. Using computer-assisted audit techniques (CAAT), the full population of journal entries is scanned for suspicious attribute values. This audit procedure is often referred to as Journal Entry Testing.
This paper proposes an approach for applying autoencoder to individual financial accounts each with a small population of journal entries. The overarching goal is to investigate the applicability and usefulness of this approach in an a realistic audit scenario. The research presented in this study follows a design science research approach.
Auto103
https://scholarspace.manoa.hawaii.edu/bitstream/10125/64408/1/0536.pdf