Deep CNN approach for Unbalanced Credit Card Fraud Detection Data

Recent developments in electronic payment technologies have significantly increased the volume of daily online transactions and payments via credit cards. As internet use grows exponentially, it naturally follows that there is also a rise in credit card fraud, which is having a big effect on many organizations, including those in the financial industry. To detect risks such as fraudulent transactions and nonuniform attacks, the development of advanced financial detection mechanisms proactively completes the required task. However, these problems have been addressed by machine learning techniques on a large scale over the past years. Hence, these techniques need some improvement in terms of identifying unfamiliar attack patterns, velocity calculations, and big data analysis. In this paper, we propose a convolutional neural network approach (CNN) along with two machine learning algorithms to tackle the issue of credit card fraud detection. Our proposed models were evaluated and compared when dealing with large amounts of data using a highly imbalanced real-world credit card fraud detection dataset. Python programming languages were used to preprocess the data and test the model's measurements and performance. As observed in the results, an accuracy of 99.7% using the Random Forest classifier was obtained, which achieved a superior result in comparison to other models.