Credit Card Fraud Detection Using Boundary Reconstruction and Integrated Classification

With the popularity of electronic payment, it not only brings great convenience, but also increases the risk of fraudulent transactions. At present, there are two problems in the identification of credit card fraud. The number of fraud and normal transactions is extremely unbalanced and the classification boundary is fuzzy. In order to solve these problems, this paper proposes an integrated classification framework for boundary reconstruction, which uses different machine learning algorithms as base learners to compare the original data with the data modeling after boundary reconstruction. The research shows that data boundary reconstruction can not only effectively alleviate the deviation caused by data imbalance to machine learning. It can also improve the data quality, so as to improve the accuracy of model classification; The integrated classification method can accurately identify credit card transactions, and the prediction effect of decision tree is the best. The proposed model is also applicable in other abnormal situations.