Financial statement fraud is a persistent challenge that undermines investor trust, corporate governance, and financial market stability. Traditional auditing approaches often fail to capture subtle manipulations within complex financial data, highlighting the need for advanced computational methods. In this study, we investigate the effectiveness of machine learning models in detecting fraudulent financial reporting. Using a publicly available dataset, we applied rigorous preprocessing, feature selection, and feature extraction techniques before evaluating five models: Logistic Regression, Support Vector Machines, Random Forest, Gradient Boosting Machines, and Deep Neural Networks. The results indicate that Gradient Boosting Machines achieved the best overall performance, with an accuracy of 94%, precision of 91%, recall of 88%, and an AUC-ROC score of 0.96. Random Forest also demonstrated strong performance, particularly in balancing recall and F1-score. These findings suggest that ensemble-based models are highly effective for identifying complex fraud patterns in financial statements. The study provides empirical evidence supporting the integration of machine learning into auditing and financial risk management systems, offering a scalable and reliable approach to strengthen fraud detection practices.
Mia et al. (Wed,) studied this question.