Leveraging Smote and Random Forest for Improved Credit Card Fraud Detection

The accelerating speed of digital transactions has made credit card theft a serious concern to customers and banks alike. Traditional means of fraud detection are no longer as strong as today's more sophisticated methods. This paper presents a summary of various machine learning techniques applied in credit card fraud detection. Several algorithms are highlighted, including logistic regression, random forests, and boosting methods such as gradient boosting and LightGBM. These algorithms were selected due to their capacity to manage big datasets and identify patterns unique to fraud. According to the results, the best performance was obtained by random forests, which achieved 99.30% accuracy, outperforming all other methods. Gradient boosting and logistic regression were also competitive, reaching 98.5% accuracy. This study offers recommendations for improving fraud detection techniques while showcasing the efficacy of machine learning in this regard.