Financial Statement Fraud Prediction System: A Deep Learning-Based Approach

Predicting financial statement fraud is a critical task for auditors and forensic accountants. Machine learning offers valuable support in this endeavor. Incorporating emerging technologies in accounting, this study employs deep learning algorithms to design a financial statement fraud detection system. Following the design science research paradigm, the study develops a framework and assesses its efficacy through prototype evaluation. The results demonstrate that the deep learning-based financial statement fraud prediction system achieves notably high prediction accuracy compared with existing methods. Moreover, the system possesses the capability of predicting specific types of fraud. The user-friendly nature of the system facilitates its adoption by auditors and forensic accountants in practical settings. Data Availability: Data are available from the public sources cited in the text. JEL Classifications: C45; C53; M41; M42.