A Combination of Deep Neural Networks and K-Nearest Neighbors for Credit Card Fraud Detection
This addresses fraud detection for financial institutions, but it is incremental as it combines existing methods on new data.
The paper tackled credit card fraud detection by combining under-sampling, K-nearest neighbors, and deep neural networks on a new dataset, achieving 98.12% accuracy in detecting fraudulent transactions.
Detection of a Fraud transaction on credit cards became one of the major problems for financial institutions, organizations and companies. As the global financial system is highly connected to non-cash transactions and online operations fraud makers invent more effective ways to access customers' finances. The main problem in credit card fraud detection is that the number of fraud transactions is significantly lower than genuine ones. The aim of the paper is to implement new techniques, which contains of under-sampling algorithms, K-nearest Neighbor Algorithm (KNN) and Deep Neural Network (KNN) on new obtained dataset. The performance evaluation showed that DNN model gives precise high accuracy (98.12%), which shows the good ability of presented method to detect fraudulent transactions.