This paper examines the growing worldwide problem of credit card fraud along with its detection techniques. This review evaluates 29 research articles published during 2019 to 2025, which focus on machine learning and deep learning techniques for fraud detection. This review is systematically organized and guided by several research questions with the objective of gaining awareness into the methodologies, data sets, and performance metrics found in the existing literature. We have thoroughly examined various factors such as from which source and which kind datasets are used, which techniques are used to deal with data imbalance problem, same way how preprocessing and feature engineering are handled, which cross validation methods are used, which feature selection methods are used and which hyper parameter tuning technique is used. We have outlined various evaluation metrics and specific machine and deep learning algorithms which are utilized in different research analysis. The analysis demonstrates a pronounced reliance on accuracy, limited engagement with cross-validation, and poor management of class imbalance. A few studies have utilized advanced feature engineering or real-time approaches. By pinpointing these shortcomings, this review establishes a framework for future research paths, such as effective evaluation metrics, hyperparameter tuning, and fraud detection based on streaming data.