Publication

Abstract : In the present day, there is a rapid shift towards cashless transactions. People started using credit and debit cards more when compared to the past. Some people are trying to take this advantage and attempting to commit fraud. In this study, the aim is to build a model that detects deceptive transactions. This paper also discusses the different strategies for handling data imbalances and the implementation of a combination of over and under-sampling methods for handling data imbalance. The work stands as evidence to understand the role of balancing the datasets and their influence on the performance of the training models. The study employed models including Decision Trees, Random Forest, Extra Trees Classifier, and XGBoost Classifier, achieving substantial accuracy, precision, and recall. Additionally, a novel ensemble learning model was proposed, using ExtraTrees Classifier, and XGBoost Classifier, resulting in even higher performance metrics. © 2024 IEEE