Developing a decision support system to detect material weaknesses in internal control

Abstract

Wells Fargo employees set up 3.5 million fraudulent accounts over several years. Wells Fargo ended up paying $4.5 billion in fines. The Wells Fargo scandal highlights the importance of management actions to prevent misstatements and potential frauds on a timely basis. This study utilized the design science research paradigm to develop a predictive framework (IT artifact) in order to stratify firms into multiple risk groups for disclosing material weakness(es) in internal control (MWIC). The proposed methodology employed a hybrid heuristic optimization-based machine learning methodology. Synthetic minority over-sampling technique (SMOTE) was utilized to handle the learning problem with the imbalanced data. The best performing model was the proposed hybrid Genetic Algorithms (GA) and Support Vector Machines (SVM). The proposed methodology was internally validated via k-fold cross validation, and then externally validated using several separate datasets. The (GA) selected variables were ranked from the most important to the least through Information fusion (IF) sensitivity. A web-based decision support system was built to predict the firm-specific MWIC risk category. The web-based tool can be used to create an early warning system for predicting MWIC(s).