Published Paper Details:

With the widespread use of credit cards, fraud has become a major problem in the credit card industry. As a result of theft, businesses and banks are reluctant to disclose the amount of money lost. Another issue with calculating credit card fraud losses is that it can only estimate the number of frauds that have been detected. It becomes harder to monitor the behavior and pattern of such transactions. Therefore, the assessment of fraud is necessary on a regular basis. The technologies like machine-learning, data-mining, and other artificial technologies are used to resolve this. In this regard, enhancing effective fraud detection using machine-learning methods is critical for empowering fraud investigators with these losses. This research paper explores how several Machine Learning models are being used to diagnose suspicious transactions. Machine learning contributes to the critical function of detecting credit card fraud during transactions. The dataset sampling methodology, selection of variables, and methods used for identification always have a massive influence on fraud detection efficiency. This system provides the necessary characteristics for monitoring both illegal and legal transactions. Classification Techniques like Decision Tree, K-nearest Neighbors algorithm, Logistic Regression, Gaussian Naive Bayes, and Artificial Neural Networks are demonstrated to identify fraudulent transactions. By training these techniques and evaluating them on various factors such as precision, accuracy, recall, and visualized the ROC curve. Based on the criteria of different methods, the best approach for detecting credit card fraud is chosen. As opposed to other algorithms, this paper shows that Decision Tree is the best optimal approach and can be used to detect further frauds