Research And Implementation Of Bank Credit Card Fraud Detection System Based On WGAN And Integrated Learning

With the rapid development of the Chinese economy,the usage of mobile network applications and electronic payment is constantly increasing,and credit card consumption is becoming more and more popular,with the scale growing larger day by day.According to the "Blue Book on the Development of China’s Bank Card Industry(2022)" data,the usage and issuance of bank cards in China are increasing year by year.Although the credit card fraud rate has decreased,card safety remains a top priority in the banking industry.Currently,credit card application fraud by commercial banks in China still exists in personal credit accounts.In recent years,there has been a growing number of research studies both domestically and internationally on credit card fraud detection.There is a considerable body of literature on integrated learning models for credit cards.This study focuses on training using foreign credit card data and data from a specific domestic bank.Due to the high dimensionality,complexity,and extreme data imbalance of the dataset,this thesis begins with data preprocessing.The minority class data of credit cards is augmented by proposing the use of a Variational Auto-Encoder(VAE)as a generator for the Wasserstein Generative Adversarial Network(WGAN).By expanding the minority class data,the quality of the generated data is ensured.Furthermore,gradient penalty is introduced to enhance the loss function,aiming to improve the convergence speed and quality of the generative model.Finally,the proposed ensemble learning algorithms,XGBoost,Random Forest,Light GBM,and Logistic Regression,are employed to classify and identify the credit card data,and the detection results are outputted.Experimental results demonstrate that compared to other methods proposed in the literature for credit card fraud detection,the approach presented in this thesis based on WGAN and ensemble learning models improves the AUC by 5 percentage points.It effectively prevents financial risks,reduces the cost of credit card fraud risks,and enhances the effectiveness of real-time risk prevention and control in financial operations.The bank credit card fraud detection system designed in this article is divided into user module and administrator module.The system is implemented based on Python language,using the Django framework,and Lay UI is used to visualize the pages.My SQL is used as the system development database.The system functional modules include credit card application,personal information management,risk result download,system management,model management,customer data analysis,and risk identification result evaluation management.Finally,the system was tested,and it was found that it effectively completed the relevant functions of each module and met the design requirements.