A distribution-preserving method for resampling combined with LightGBM-LSTM for sequence-wise fraud detection in credit card transactions

Behnam Yousefimehr, Mehdi Ghatee

Abstract: Highlights•A resampling method to preserve the distribution of fraud instances.•Combining OCSVM with SMOTE and random undersampling to detect outliers.•Light Gradient-Boosting Machine (LightGBM) and Long Short-Term Memory (LSTM).•Determining the optimal window size for LSTM to analyze sequence-wise fraud.•Getting F1 score of 87% and AUC score of 96% on European credit cards fraud.