A Machine Learning Approach for Credit Card Fraud Detection using Feature Engineering and Ensemble Models

Credit card fraud remains a major concern in the digital economy of today, posing serious financial and security risks. This study presents a comprehensive approach to fraud detection by leveraging machine learning, specifically the Random Forest Classifier, to accurately identify fraudulent credit card transactions. A publicly available data set was used and extensive data preprocessing steps were performed, including feature selection, handling missing values, and addressing class imbalance through undersampling. The model was trained to distinguish between fraudulent and legitimate transactions, achieving a precision of 98. 7%, with a precision of 97. 3%, a recall of 96. 8%, and an AUC-ROC score of 0.99. A detailed evaluation using the confusion matrix, classification report, and ROC curve validated the effectiveness of the model. Furthermore, analysis of the importance of the features identified the amount and location of the transaction as critical indicators of fraud. The results demonstrate that the Random Forest Classifier is a robust and reliable method of fraud detection. This comprehensive study offers actionable insights and practical implications for the development of more secure and efficient fraud detection systems in financial services.