The Efficacy of Financial Ratios for Fraud Detection Using Self Organising Maps

Abstract

In recent times, financial statement fraud has resulted in billions of dollars being lost from the financial system. Financial statement fraud is a problem for both listed and local government entities. The present focus in the literature has been on analysing listed entities, and the analysis is typically framed as a supervised learning problem with the labels being audit opinions. In this paper we assess the efficacy of using financial ratios for detecting fraud in financial statements of local government entities. The problem is framed as an unsupervised learning problem. Self organising maps are used due to their visual nature and the resulting accessibility of information to decision makers. The analysis shows that financial ratios are useful in the detection of fraud in the public sector. Using qualified audit opinions as an indication of fraud, the analysis shows that a high current ratio is associated with entities that have unqualified audits (i.e. non-fraudulent), while entities that are fraudulent have a high debt to revenue ratio.