On Credit Card Fraud Detection Using Machine Learning Techniques

Payment through credit card is widely used in many applications. On one hand credit car-based payment enables user convenience and flexibility in electronic business; on the other hand, detection of credit card frauds has become an important concern in the card industry, which leads to a colossal amount of loss of financial institutions and the card users. In this work, we discuss four different machine learning models (Decision Tree, Random Forest, Isolation Forest and Support Vector Machine) to analyze credit card frauds with respect to accuracy, recall and precision and F-score metrics using Kaggle dataset. The experimental results show Random Forest performs better among others for credit card fraud detection. Furthermore, it is observed that if an organization wants to focus on detecting frauds and ignores the false alarm of frauds, then Isolation Forest can be used to detect fraudulent transactions.