Fraud detection based on FS-SMOTE model for credit card

Abstract

In the financial security technology, credit card fraud detection technology is an important technical means, which collects and analyzes the credit card transaction data in a certain period of time, detects fraud in many credit card transactions, and takes the corresponding alarm response. At the same time, in view of the extremely imbalanced characteristics of credit card fraud customer data set, the number of minority samples is increased by Smote, which is a representative algorithm of oversampling technology. Logistic regression, KNN, Decision tree, Bagging and Stochastic gradient descent are used to construct fraud detection models. The credit card fraud data released on Kaggle platform were selected for verification. The experimental results show that the fusion model based on Feature selection and Smote greatly improves the accuracy and detection rate of credit card fraud detection as a whole.