Research On Credit Card Fraud Detection Based On Random Forest

With the growth and widespread of the internet the credit card payment industry has expanded rapidly during the course of time.And credit card becomes the most popular mode of payment for both online as well as regular purchased,cases of fraud associated with it are also rising.In real life,fraudulent transactions are scattered with genuine transactions and simple identification techniques are not often sufficient to detect those frauds accurately.In order to ensure the security of credit card payments,it is particularly important to use artificial intelligence technology to detect fraud in credit card transactions.Credit card transaction data has problems such as large data volume,unbalanced dataset,large computational complexity,and low recognition rate,In this thesis,by studying the imbalanced data classification methods,credit card data characteristics and random forest algorithm,an improved treatment of the problem of imbalanced data classification random forests algorithm is proposed.It first oversamples the credit card data,secondly reduces the dimension of the training sample,and finally uses random forest for fraud detection and identification.Through the analysis of the characteristics of credit card data,a method of clustering and selecting more representative positive samples for overfitting is proposed,which effectively solves the problem of high false positive rate in the classification results of credit card data sets,and Experiments show that this overfitting method is also applicable to financial data with similar characteristics to credit card data sets.Regarding the problem of low classifier accuracy,a selection method for classifier decision trees is proposed,which effectively improves the accuracy of the random forest algorithm.In this thesis,through a large number of experiments,the parameter selection and performance indicators of credit card fraud methods are analyzed,which has important academic significance and practical value for the use of random forests to solve the classification of imbalanced data sets.