Research On Credit Card Fraud Detection And Recognition Based On GAN-CNN Unbalanced Classification Algorithm

With the development of the Internet in recent years,payment methods have become increasingly diversified.Credit card payment,as one of the methods,has increasingly become the norm in life.As more and more people use credit cards to purchase products they need in life,the illegal cases of credit card fraud are increasing year by year,and the economic loss caused by credit card crimes has reached tens of billions every year.Credit card fraud has not only greatly disrupted The financial order has also endangered the healthy development of the credit card industry.In order to solve this problem,in addition to the relevant laws and regulations promulgated by the state to protect people's rights and interests,financial institutions also need to implement credit card fraud control measures.Therefore,credit card fraud detection technology has become the top priority of the financial management system.The core issues to be discussed in this essayIn the past,credit card fraud detection algorithms usually involve data mining and machine learning-related technologies,such as decision trees,XGBOOST,Random Forest,GBDT,Logistic Regression,SVM,and Bayeux Sri Lanka network,neural network,etc.,these are common algorithms used to solve two classification problems.However,they have extremely stringent requirements for the training data,that is,not only the sample data of normal transactions in the data set,but also the sample data of abnormal transactions with the same amount of data,that is,fraud data.However,in real application scenarios,credit card fraud transactions account for a very low proportion of total transactions.If the data is directly trained without processing,the effect is very poor.Therefore,it is necessary to solve the imbalanced binary classification problem of credit card fraud and improve the accuracy of small samples.The recognition rate is the focus of this thesis.For credit card fraud detection scenarios,this thesis attempts to use the popular deep learning methods to solve such problems,and proposes a new algorithm that combines Generative Adversarial Networks(GAN)and Convolutional Neural Networks(CNN),that is,the GAN-CNN algorithm model.Credit card fraud is an imbalanced binary classification problem.This algorithm uses GAN to generate a small number of fraud samples to balance the data set,and then the balanced processed data set is trained by the CNN algorithm.The CNN algorithm can continuously optimize the model and perform the input training set.Feature learning until more accurate classification is achieved.Experiments have shown that using the GAN-CNN algorithm to detect unbalanced credit card fraud samples can achieve better detection results than other methods.