esearch on Credit Card Fraud Classification Based on GA-SVM

Credit card transaction data belongs to imbalanced data. It is necessary to identify whether the credit card is at risk of fraud by analyzing the previous transaction data, so as to deal with the risk in time. Aiming at the problem that a traditional classifier cannot achieve good results in dealing with imbalanced data classification, a Support Vector Machine (SVM) model optimized by Genetic Algorithm (GA) is proposed in this paper. First, cluster centroids sampling is used to make the dataset relatively balanced. Second, we employ GA to optimize the parameters of the SVM and select optimal features of the data to improve the classification performance. The Area Under Receiver Operating Characteristic (ROC) Curve (AUC) and accuracy serve as classification evaluation indicators. Compared with traditional classifiers, accuracy and AUC value have been improved by GA-SVM. Compared to Random Forest, Logistic Regression and Naive Bayes classification, GA-SVM has improved the accuracy by 2.03%, 5.14% and 6.89%, respectively. On the AUC value, it is increased by 2.00%, 5.36% and 6.93%, respectively. These results show that GA-SVM can improve the overall accuracy and optimize the performance of the classifier on the credit card fraud recognition.