An Efficient Credit Card Fraud Detection Using SMOTE Under Machine Learning Environment

Credit card fraud detection refers to the process of identifying fraudulent transactions made using a credit card. Credit card fraud can cause significant financial loss to both cardholders and financial institutions. If fraudulent transactions are not detected and stopped promptly, they can result in unauthorized charges and potentially lead to identity theft or credit card fraud rings. Therefore, to stop these frauds we need an important fraud detection system that detects them directly. This paper gives a comparative study before and after applying the SMOTE technique. In this paper, we’ve applied machine learning algorithms on an imbalanced dataset such as a Random Forest Classifier, CatBoost Classifier, and Decision Tree using the Synthetic Minority Oversampling Technique (SMOTE). So these models are applied to the data and the model performance is estimated based on quantitative measures such as accuracy and AUC-ROC score. The research shows that using SMOTE accuracy of the Random Forest Classifier is 0.9996, CatBoost Classifier is 0.9979 and for the Decision Tree the accuracy is 0.9926.