Credit card fraud detecting using machine learning classifiers in stacking ensemble technique
Muhammad Daud Abbasi, M. A. Shah · IET conference proceedings. · 2022
Credit cards have been a primary target for scammers across the globe due to the low risk and high reward nature of this type of fraud. Credit card fraud detection datasets are mostly unbalanced because, in day-to-day life, only a handful of transactions are fraudulent. Most credit card transactions are normal and the fraudulent ones are very similar to normal transactions. In our paper we describe how we designed a model to work on unbalanced datasets by using an ensemble technique of stacking classifiers; we chose RF, XGB, KNN and MLP to be the classifiers in level 0. Level 0 prediction will come to the level 1 meta classifier. The meta classifier in our model was logistic regression. By applying different experiments to this stacking classifier, we achieved a model that gives more accurate results than those shown in previous literature.