This paper discusses the growing worry about credit card theft in the context of increased internet purchases. With a focus on real-world scenarios, the study uses a variety of machine learning models to target four major forms of fraud, to assist banks and financial institutions in picking the best algorithms for certain fraud categories. The project emphasises real-time detection by utilising predictive analytics created by installed machine learning models and an API module, allowing for rapid assessment of transaction authenticity or fraud. An important component is dealing with sketched data distribution, and the research provides a unique technique to manage this issue efficiently. The experimental data for the study are gained from a financial body under a secret disclosure bond, protecting secrecy and security of the information.