Garra Rufa Fish Optimization-based K-Nearest Neighbor for Credit Card Fraud Detection

Credit cards are one of the popular modes of payment in the online sector, and the activity of fraudulent utilizing credit card payment technology is increasing rapidly. A significant number of these transactions are debit or credit card transactions, as a deployment of digital transactions maximized, fraudsters also enhanced on corresponding platforms. However, it is a difficult task for financial institutions, banking organizations, and financial technology firms. In this research, the Garra Rufa Fish Optimization-based K-Nearest Neighbor (GRFO-KNN) is proposed for Credit Card Fraud Detection (CCFD) using Machine Learning (ML). Initially, the data is gathered from a CCF dataset to evaluate the proposed method. The z-score normalization is employed for measuring input dataset and Synthetic Minority Oversampling Technique (SMOTE) is established to balance the imbalanced data. Then, the GRFO is employed to select the appropriate features and KNN is performed to detect and classify the CCFD effectively as fraud or legitimate. The existing methods like one-dimensional Dilated Convolutional Neural Network (DCNN), AllK-Nearest Neighbors undersampling approach with CatBoost (AllKNN- CatBoost), and Long Short-Term Memory (LSTM)-attention technique are compared with GRFO-KNN. The proposed GRFO-KNN achieves a better accuracy of 0.9998 compared to one-dimensional DCNN, AllKNN-CatBoost, and LSTM-attention respectively.