Abstract

In the face of a financial security point of view fraud detection systems emerge as key element in as far as the threats of losing money as a cost are concerned. The act of resistance against impersonation in the financial transactions arena has become very important in terms of preserving privacy. This study involves discussion of application of Machine Learning algorithms which includes Decision Tree, Random Forest and XGBoost for the process of detection of fraudulent credit card transactions. These models were selected because they can handle large and complicated data sets and derive hidden patterns that represent fraudulent activity. By making use of such features as transaction history and user behavior, the proposed system was able to show off its exceptional accuracy in differentiating normal or fraudulent transactions. The result for Random Forest model is 99.96% accuracy, the result for Decision tree model is 99.92% and XG boost model result accuracy is 99.95%. These results validate machine learning to continuously detect fraud in financial services. Moreover, continuous retraining of models is important to adapt to changing fraud tactics to make the system effective and scalable in the long term.