Descripción:
Fraud detection in financial statements is a constant and laborious task in the audit area. Traditionally, this task has been performed by experts, limiting its scope due to restrictions in manual processing capacity. In recent years, there has been an increase in the use of data mining and machine learning techniques to review in a comprehensive and automated way the organizations’ financial statements. The objective of this study was to analyze data mining and machine learning techniques used in financial fraud detection, in order to characterize the reported algorithms and the metrics used to evaluate their effectiveness. For this, a systematic mapping study of 67 studies was carried out. Our results show that since 2015 there was an upturn in the amount of studies that use these techniques for fraud detection in financial statements, where vector support machines are the most used technique, with 19 studies, followed by artificial neural networks, with 15 studies, and decision trees, with 11 studies. Effectiveness was assessed by the degree of precision with which the implemented techniques detected real fraud cases, obtaining values between 70% and 99.9%.
Tipo de publicación: Journal Article
Publicado en: Revista Ibérica de Sistemas e Tecnologias de Informação
Autores- Ramírez-Alpízar,Alex
- Jenkins,Marcelo
- Martínez,Alexandra
- Quesada-López,Christian
Investigadores del CITIC asociados a la publicación
Dr. Christian Quesada-López
Dra. Alexandra Martínez Porras
Dr. Marcelo Jenkins Coronas
Proyecto asociado a la publicación
Proyecto sombrilla
- Algorithms
- Certification testing
- Computers–Information Science And Information Theory
- Data mining
- Decision trees
- Financial statements
- Fraud
- Kernels
- Machine learning
- Mapping