From Conventional Methods to Contemporary Neural Network Approaches: Financial Fraud Detection

Abstract

This chapter provides insights on the underlying reasons to replace the conventional methods with contemporary approaches—the neural network-based machine learning methods—in financial fraud detection. To do this, we perform a systematic literature review on the evolution of financial fraud detection literature over the years from traditional techniques toward more advanced approaches such as modern machine learning methods like artificial neural networks. Additionally, this chapter provides concise chronological progress of the fraud literature and country-specific fraud-related regulations to draw a better framework and give the idea behind the corpus. Using the metadata in the existing literature, we show both benefits and costs of using machine learning-based methods in financial fraud detection. An accurate prediction using contemporary approaches is essential to minimize the potential costs of fraudulent financial activities for stakeholders, reduce the adverse effects of fraudsters’ and companies’ fraudulent activities, and increase trust in capital markets via continuous fraud risk assessment of companies.