This paper reflects an In-Depth comprehensive comparative study of fraud in credit card leveraging the XGBoost Algorithm. Because fraudulent transactions can result in tremendous financial losses, credit card fraud detection is a critical problem for financial organizations around the world. In this paper, the main focus of the research work is on the XGBoost algorithm, and hence this paper presents a detailed review of machine learning approaches for credit card fraud detection. We take a publicly available, anonymized credit card transaction dataset and try to overcome the class imbalance problem and improve our model's performance using feature engineering and data preparation methods. We used various machine learning models such as Random Forest, Support Vector Machines, and Logistic Regression in combinations with XGBoost to check whether the accuracy and classifier robustness were significantly higher in classification tasks. Models are evaluated based on how accurate, precise, or sensitive they are about their F1-score. Our finding is that the performance of the XGBoost model in identifying fraud transactions is better than others. Hence, it is more practical in real life. The study's conclusions aid in the creation of more trustworthy fraud detection tools to reduce financial risks in the banking industry.