Credit card fraud detection using XGBoost for imbalanced data set
Archana Purwar, Manju Manju · 2023
Frauds using credit card are easily done by the fraudsters. Due to increase in fraud rates all over the world, various machine learning algorithms are being used by analysts and researchers to detect and analyze frauds in online transactions. However, training data set may have a few instances of one and more instances of another class in case of binary classification particular class which makes result biased. Hence, this paper sets an objective to give methodology which is able to detect fraud accurately in case for skewness of data. The proposed method compares different techniques to handle imbalance problem and chooses best approach out of these and uses XGBoost as classifier to predict whether transaction is fraudulent or not. The developed method is evaluated using European credit card fraud dataset and obtained better F1 score, recall and accuracy as 82.78%, 78.9% and 99.3% respectively as compared to other algorithms taken under study.