Credit card fraud identification based on unbalanced data set based on fusion model

With the rapid development of bank credit card business, banks are facing increasingly serious credit card fraud problems. In order to carry out the credit card business and effectively prevent fraud risks, this paper analyzes the research results at home and abroad, comprehensively considers the availability, security and timeliness of the model. For the unbalanced data set, using machine learning methods, three kinds of anti-balance are proposed. Fraud model. Firstly, data preprocessing is performed by using undersampling method. Secondly, Lasso-Logistic, XGBoost, and the above two models are used to model and predict. Finally, the three methods are compared and optimized. This paper analyzes fraudulent behavior through three kinds of anti-fraud systems, which has positive and important significance for improving the accuracy and effectiveness of credit card fraud detection.