Visual Exploration of Journal Entries to Detect Accounting Irregularities and Fraud

Nowadays, companies and organizations register millions of accounting transactions each year. Although most of these journal entries are legitimate, auditors face legal and financial obligations to discover transactions that are fraudulent. In this work, we present a visual analytics workflow to quickly identify unusual transactions in accounting data. In a first step features are derived from journal entries and are clustered to identify transactional patterns. In a second step the data is visualized to support the identification and investigation of unusual transactions. Following this workflow auditors are given the chance to identify new suspicious transactions that might correspond to fraud. We evaluated the proposed approach in a real world scenario by analyzing accounting data and discussed the results with auditors.