Improved credit card fraud detection method based on XGBoost algorithm

With the development of Internet and technology, credit cards are more widely used transaction data larger. The set card fraud is a typical imbalanced problem. model should ensure that detected customer service quality guaranteed. Improving both precision recall rate focus current research. However, when constrained by level machine learning, it good choice to use different models evaluate obtain for manual use. Few papers results comprehensive judgment. In this paper, two sampling methods (undersampling-NearMiss, oversampling-SMOTE) three algorithms (Logistic, Neural Network, XGBoost) analyze based on records in Europe within days September 2013. above six cases were compared explore appropriate detection methods. study found Nearmiss-XGBoost had best rate, SMOTE -XGBoost result precision. Compounding can improve recall.