Fraudulent Credit Card Transactions Prediction Using 2D-Convolutional Neural Network

Abstract: Fraudulent credit card is one of the most alarming concerns in this modern age. With few misuse of credit card, thousands of dollars can be mishandled. This paper focused on detecting fraud credit card transaction using 2D- Convolutional neural network. The paper reports the categorization of two designated categories- Genuine and fraud transactions. In the pre-processing stage we applied under-sampling technique to handle imbalance dataset. For the improvement of the accuracy, we have decreased the number of convolutional layers with a sigmoid layer at the top. The proposed technique achieved an accuracy of 94%

Keywords: Credit card fraud detection, convolutional neural network, under-sampling, imbalanced dataset.