Credit card fraud detection using supervised machine learning methods

Abstract

Nowadays, people are using credit cards more frequently, especially for online shopping. Credit card fraud has a negative influence on individuals, merchants, and financial institutions. This paper focuses on credit card fraud detection by using three supervised machine learning methods: decision tree, random forest, and AdaBoost algorithm. These three methods are widely used in the medical field, chemical field, visual identification, and finance field. We use the confusion matrix and receiver operating characteristic curve to interpret our model. The random forest has the highest accuracy of 99.6% in predicting non-fraud. The AdaBoost has the highest accuracy of 77% in predicting fraud transitions, while it has the lowest accuracy of 95% in predicting non-fraud. Therefore, the random forest algorithm is the best model to apply in credit card fraud detection.