Transaction fraud detection by CatBoost model with feature engineering
Jiahan Shi · Applied and Computational Engineering · 2024
In the digital era, the ubiquitous nature of online transactions has placed a spotlight on the imperative for robust fraud detection systems. This research tackles this pressing issue, with a particular emphasis on the nuanced process of feature engineering tailored for the IEEE-CIS dataset, a representative sample of contemporary transactional behaviors. The enhancement of data attributes, through meticulous feature engineering, acts as the bedrock for the application of the CatBoost model - a gradient-boosting technique revered for its precision. The paper dedicates significant attention to both these pivotal stages, showcasing the synergistic effect they have when applied in tandem. With this refined data and sophisticated modeling, the proposed method manifests exceptional performance, establishing a new benchmark when juxtaposed with other gradient-boosting methodologies. Conclusively, this study offers valuable insights for enhancing online transaction security and sets the stage for further innovation in fraud detection within the machine learning community.