۱- Introduction

۲- Design of a fraud detection process for financial statements of business groups

۳- Development of fraud detection techniques for financial statements of business groups

۴- Demonstration and evaluation of the proposed approach

۵- Conclusions

References

Abstract

Investors rely on companies’ financial statements and economic data to inform their investment decisions. However, many businesses manipulate financial statements to raise more capital from investors and financial institutions, which reduces the practicality of financial statements. The modern business environment is highly information-oriented, and firms’ information systems and activities are complex and dynamic. Technology for avoiding fraud detection is continually updated. Recent studies have focused on detecting financial statement fraud within a single business, but not within a business group. Development of methods for using diverse data to detect financial statement fraud in business groups is thus a high priority in the advancement of fraud detection.
This study develops an approach for detecting fraud in the financial statements of business groups. The proposed approach is applied to reduce investment losses and risks and enhance investment benefits for investors and creditors. The study objectives are achieved through the following steps: (i) design of a process for detecting fraud in the financial statements of business groups, (ii) development of fraud detection techniques for use with such statements, and (iii) demonstration and evaluation of the proposed approach.

Introduction

Since the Procomp case and Enron event, investors, governments, and regulatory authorities have begun to focus on financial statement fraud committed by business groups. Falsified financial statements may result in large losses for investors and creditors in capital markets. The modern business environment is highly information-oriented, and firms’ systems and activities are complex and dynamic. Technology used to avoid fraud detection is constantly updated. Development of methods for using diverse data to detect financial statement fraud in business groups is thus a high priority in the advancement of fraud detection. Various approaches have been developed to detect fraud in corporate financial statements. Kirkos et al. (2007) explored the effectiveness of data mining (DM) classification techniques for detecting firms’ fraudulent financial statements (FFS) and identified factors associated with FFS. Auditors assisted in detecting fraud using DM techniques. The study also investigated the usefulness of decision trees, neural networks, and Bayesian belief networks in identifying FFS. Ravisankar et al. (2011) also used DM techniques such as multilayer feed forward neural network (MLFF), support vector machine (SVM), genetic programming (GP), group method of data handling (GMDH), logistic regression (LR), and probabilistic neural network (PNN) to identify companies that had committed financial statement fraud. Each of these techniques was tested on a dataset covering 202 Chinese companies, and the results of tests with and without feature selection were compared. Among the techniques, PNN was the most accurate without feature selection, and GP and PNN were the most accurate with feature selection (with marginally equal accuracies).