Predicting Financial Statement Fraud using Artificial Neural Networks

Abstract

Financial statement fraud is ever-increasing within companies and is stirring chaos worldwide. The issue lies within a misguided tone at the top, which makes it difficult to identify fraud and takes a longer time to uncover. The costs associated with the deliberate management misrepresentation could lead to insolvency and affect the company itself, investors, stakeholders, as well as the economy as a whole. This research aims to construct a predictive model using a key technology used in the financial field, artificial neural networks (ANNs). The study uses data from 50 fraudulent companies and 150 non-fraudulent companies from the U.S. SEC, NYSE, LSE and Athex, where variables are computed based on the fraud risk indicators from ISA 240. A multi-layer perceptron feed-forward neural network with a back-propagation algorithm was utilized to construct the model. The results show the predictive accuracy of the ANN model at 93.3%.