Abstract: Due to the increased use of credit cards for online purchases, fraud has increased exponentially because of the rapid rise in the e-commerce business. In recent years, banks have found it increasingly difficult to detect credit card fraud. The power of machine intelligence can detect credit card fraud. Banks have used a variety of machine learning approaches, prior data and novel attributes to better forecast these transactions. For credit card transactions, the sampling method used, as well as the selection of data points and the detection techniques used, can all have a significant impact on fraud detection. Credit card fraud is investigated using logistic regression, decision trees, random forests and Support Vector Machines (SVM). In September 2013, the data included all transactions made by European cardholders. There were 492 instances of fraud out of a total of 284, 807 transactions. It classifies fraudulent transactions as "positive" and legitimate transactions as "negative. Fraud accounts for 0.173% of the total transactions in the data set, making it highly imbalanced. To balance the data set, over sampling was used, resulting in 60% of fraudulent transactions and 40% of genuine transactions. Among the four models, Logistic Regression produced the best results, the Logistic Regression model has a 96% accuracy.
Sameh Gamal Khalil Taktak, Atef Zaki Ghalwash, Amr Galal and Mohamed M. Abbaassy, 2022. Machine Learning for Credit Card Fraud Detection. Asian Journal of Information Technology, 21: 6-10.