Accounting and Finance

Unmasking fraudulent financial reporting

Supervisor

Prof. Kevin Dow, Dr Dulani Jayasuriya
Faculty of Business
Project code: BUS001

This research project aims to unravel the underlying behavior of firms that have committed financial statement fraud. We extend existing fraud research by developing a predictive model that can be used by auditors and regulators. We will employ a predictive analytics/machine learning approach to properly classify firms according to the likelihood of engaging in financial statement misrepresentation. We will examine whether a firm’s fraudulent behavior is more likely when their net income is close to the motivated triggers. We will implement several machine learning algorithms including Support Vector Machines (SVM), Random Forests (RF) and other classification algorithms for this purpose. We believe machine learning methods may classify frauds with higher accuracy and therefore provide better insights relative to traditional methods. Moreover, we intend to compile a unique database for fraudulent reporting which could be used for further machine learning based fraud detection projects in collaboration with industry partners.