Detecção de fraudes em transações financeiras com Machine Learning: uma revisão da literatura
The growing digitalization of financial transactions has significantly increased the occurrence of fraud, making its detection a central challenge for banking institutions and payment companies. Given the limitations of traditional rule-based systems, machine learning has gained a prominent position in recent scientific literature. This study aimed to analyze, through a narrative literature review, the main machine learning techniques applied to fraud detection in financial transactions, with emphasis on credit cards. The research, qualitative and exploratory in nature, was based on the critical analysis of eight scientific works published between 2018 and 2026, found in databases such as IEEE Xplore, Scopus, ScienceDirect, and Google Scholar. Results indicate the recurrent use of ensemble algorithms, such as Random Forest and boosting, alongside classical models, such as Logistic Regression, Decision Trees, and Support Vector Machine, with no universally superior algorithm. The extreme data imbalance is addressed through undersampling, SMOTE and its variants, and decision threshold adjustment. It was also found that metrics such as F1-Score, AUC-ROC, and the area under the precision-recall curve are more suitable than accuracy alone, and the tension between opaque models and the right to review automated decisions set out in the Brazilian General Data Protection Law was discussed. It is concluded that gaps remain regarding model interpretability, adaptation to emerging fraud patterns, and the scarcity of real and public data for validation, pointing to relevant directions for future research.
Authors
- Fábio Almeida do Nascimento
- João Paulo de Assis Marques
- Alexandre Siqueira Souza Costa
Publication Details
- Journal
- Revista fisio&terapia.
- Published
- 2026-11-15
- DOI
- https://doi.org/10.69849/6va6aj45
- Primary Topic
- Imbalanced Data Classification Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00