Credit card fraud poses a significant challenge for banks, as conventional fraud detection methods often fall short in identifying fraudulent transactions. This paper introduces a credit card fraud prediction system that harnesses the power of machine learning abnormality and regression algorithms, complemented by a user-friendly web application. The system aims to forecast fraudulent transactions by analysing historical data. Python programming language, along with widely adopted open-source libraries such as Scikit-learn, Pandas, and Flask, forms the foundation of its implementation. The web application enables users to interact with the system, input transaction data, and obtain predictions. Empirical findings substantiate the system's exceptional accuracy in detecting fraudulent transactions, rendering it a valuable asset for banks seeking to fortify their fraud detection capabilities. Additionally, the system's adaptability allows for customization, facilitating integration with other systems and the integration of supplementary security features. In conclusion, the credit card fraud prediction system proposed in this paper offers an efficient and precise solution for banks combatting credit card fraud.