Credit cards are convenient and simple to use, so consumers might use them for online purchases. Credit card abuse has increased in parallel with expanding credit card usage. Through a focus on frauds with a high rate of false alarms, the utilization of publicly available data, a significant class imbalance, and a variety of fraud types, this research study seeks to discover credit card theft. Reducing the amount of money that customers and financial institutions lose is the aim. Numerous machine learning methods, such as XG Boost, Decision Tree, Random Forest, Support Vector Machine, Logistic Regression, and Extreme Learning Method, are discussed in the literature when it comes to credit card identification. To minimize losses from fraud, more advanced DL techniques are needed because current algorithms are not particularly successful at spotting fraud. Using the most recent advancements in DL algorithms for this purpose is the primary goal. To determine the best fraud detection strategy, the researchers compared ML and DL approaches in a study. They carried out their investigation using the European card benchmark dataset. Initially, they used a machine learning technique to enhance fraud detection to a certain extent. Later, they used three distinct convolutional neural network models to improve fraud detection efficacy. Additional To increase detection accuracy, the researchers used more layers. They made adjustments to hidden layers and epochs, among other aspects, and employed sophisticated models in their analysis. Significant improvements were observed, as evidenced by the accuracy of 99.9% and the optimal values of other measures like precision, f1-score, and AUC curves. The proposed model outperforms current ML and DL approaches for credit card recognition tasks. We experimented with data balancing strategies and used DL approaches to lower the amount of mistakenly recognized negatives. In real life, credit card theft can be detected with the help of the tactics being described.