<![CDATA[In today's digital economy, credit cards are essential and both credit card usage and theft have increased significantly in recent years. Credit card fraud can be categorized using machine learning models using data from suspicious transaction history. However, credit data is often imbalanced. Therefore, machine learning models are biased towards the majority class resulting in poor performance on publicly accessible Kaggle credit card classification datasets. We balance the class distribution in the dataset using a hybrid synthetic minority oversampling strategy to address this difficulty. The findings show that the random forest machine learning model combined with oversampling techniques combined with feature engineering and cross-validation yields optimal results of more than 99% for all assessment measures. It performs better compared to three other models, namely decision tree, gradient boosting, and XGBoost. It can be concluded that the use of feature engineering, cross-validation, and oversampling are useful approaches to handle imbalanced credit card data and ultimately help in preventing credit card transaction fraud.]]>
Copyrights © 2024
