Classifier Ensembles for Credit Card Fraud Detection

Abstract

There is a risk of payment card abuse in e-commerce, so it is important to note those transactions that are took place without the owner's knowledge. We use various predictive models to see how accurate they are in detecting whether a transaction is a normal payment or a fraud. In classification problem, Bagging, AdaBoost, Random forest and Gradient boosting classifier ensembles are used. We chose decision trees as the base classifiers in classifier ensembles because they are very accurate and sensitive to rotation of the feature axes. Consequences of employing different classifier ensembles are monitored. We present comparisons of classifier ensembles with different performance measures.