Digitalni repozitorijum
Univerziteta u Beogradu
{{ mc.sd.lang | uppercase }}
APPLICATION OF MACHINE LEARNING METHODS TO BENFORD’S LAW ANALYSIS: POTENTIALS, AND LIMITATIONS
Stanojević, Jelena
Radojičić, Dragana
Azdejković, Dragan
Abstract: Benford’s Law has long served as a valuable statistical tool for detecting anomalies in numerical datasets, particularly in fraud detection, auditing, and forensic accounting. Traditionally, its application relies on hypothesis testing methods such as the Chi-square test, Kolmogorov–Smirnov test, or the Mean Absolute Deviation. Recent years have seen growing interest in enhancing Benford’s Law-based analyses through the application of machine learning (ML) techniques, particularly in contexts where traditional statistical methods face limitations. This study surveys the current state of the art at the intersection of Benford’s Law and machine learning, examining both supervised and unsupervised approaches for detecting deviations from expected digit distributions. We also review methodological advances across diverse domains such as financial auditing, digital forensics, and fraud detection, and outline promising directions for future research and applications. Special emphasis is devoted to future research, addressing its possible directions and limitations, along with data-related and methodological challenges. We conclude by outlining promising avenues for further study and practical implementation.
These Conference Proceedings have been funded by the Bureau of International Narcotics and Law Enforcement Affairs (INL) of the United States, within the framework of the project “Building Forensic Accounting Capacities in Serbia.”
engleski
2025
Ovo delo je licencirano pod uslovima licence
Creative Commons CC BY-NC-ND 4.0 - Creative Commons Autorstvo - Nekomercijalno - Bez prerada 4.0 International License.
http://creativecommons.org/licenses/by-nc-nd/4.0/legalcode
Key words: Benford’s Law, anomalies, forensic accounting, auditing, machine learning (ML) methods, anomaly detection algorithms