Application Of Machine Learning In Credit Card Fraud Detection

In recent years,China's credit card business has developed rapidly,and the issuance volume has increased year by year.Credit cards have played a huge role in stimulating consumption and stimulating economic growth.It also brought huge contributions to domestic economic development.However,with the increase in the use of credit cards,credit card frauds occur frequently.Common credit card frauds include:1.Fraud using forged or invalid credit cards;2.Fraud using other people's credit cards;3.Malicious overdraft of credit cards.As more consumers start online transactions,credit card fraud detection faces huge difficultiesThis paper uses Vesta's credit card transaction data set and uses machine learning algorithms to establish a credit card fraud detection model.Since the data set contains more features,a method based on random forest feature screening is used to retain 50 important features for modeling.Aiming at the problem of data imbalance,SMOTE technology is used to make the ratio of positive and negative samples reach 1:1.For the preprocessed data set,a variety of machine learning methods are used for modeling and analysis,including logistic regression,support vector machines,and LightGBM models,and the models are fused through the Stacking method.LightGBM performed well in the credit card fraud detection data,with AUC reaching 0.918.Compared with a single model,the model after stacking fusion has a certain improvement,and the AUC has increased to 0.925.Both LightGBM and Stacking integration models can be more effectively applied to card fraud detection.