A Combined Support Vector Data DescriptionFuzzy Logic Approach for Anomaly Detection in Credit Card Transactions

Now a days, as the technology has revolutionized a number of online transactions which had grown up in large sizes and grabs a vast portion of these transactions. The speedy involvement of credit card transactions resulted in tremendous losses to the people’s wealth as increase in the fraudulent transactions. The traditional models which are assisted in detection of anomalies had less accuracy and failed in classifying complete anomalies. Therefore, this paper proposed a Support Vector Data Description (SVDD)-Fuzzy Logic for detection of anomalies in credit card transactions. The proposed method employed on credit card fraud detection dataset which holds transaction details of European Credit Card (ECC) holders. The transaction details obtained from dataset that has balanced by Synthetic Minority Oversampling Technique-Edited Nearest Neighbour (SMOTE-ENN) technique. These balanced data features are selected by Recursive Feature Elimination (RFE) method which reduced the least significant features for selecting appropriate features. The obtained relevant features had classified into anomalies and positive datapoints by SVDD which employed Polynomial Self Learning Particle Swarm Optimization (PSLPSO) and Fuzzy Logic for improving model accuracy. The proposed SVDD-Fuzzy Logic achieved better accuracy ($98.50 \%$), Recall ($95.25 \%$), precision ($98.90 \%$) and F1 Score ($97.45 \%$) compared to existing Generative Adversial Network.