The proposed dissertation research will use machine learning to develop a risk-assessment algorithm to aid in corporate financial statements and reporting fraud (FSRF) detection. The project will use SEC Audit and Enforcement Releases (AAERs) data from 2002-2018 and corresponding SEC prosecution data. Fraudulent firms will be matched with non-fraudulent firms using SEC annual reports (10-Ks) to assist in identification of risk factors.
The applicant proposes to use random forest (to model the complex decision-making process underlying fraudulent reports) and deep network (to model unobservable processes) machine learning techniques to train and test the risk assessment algorithm. These methods are superior to other commonly used statistical methods that have difficulty with non-linear relationships, smaller degrees of freedom, and multi-collinearity. Finally, the algorithm will be made available to criminal and regulatory agencies through targeted research briefs and publication in trade journals.
"Note: This project contains a research and/or development component, as defined in applicable law," and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14). CA/NCF