The widespread adoption of the Unified Payments Interface (UPI) has simplified digital transactions, enabling fast and easy payments. However, with its increasing usage comes a surge in fraudulent activities. This research paper focuses on detecting fraudulent activities using machine learning algorithms. The study is conducted on UPI fraud detection using XGBoost (Extreme Gradient Boosting) and Random Forest machine learning algorithms. To test the algorithms, an online payment fraud detection dataset from Kaggle is used. The effectiveness of the results is analyzed using Accuracy, Precision, F1-Score, Macro Average, and Weighted Macro Average. The results indicate that XGBoost performs better on smaller datasets, while Random Forest performs better on larger datasets. To achieve higher accuracy, an ensemble model combining XGBoost and Random Forest is implemented using logistic regression. This ensemble approach yields higher accuracy for both smaller and larger datasets.