Using neural network for credit card fraud detection

Abstract

In the present time, the internet has become more popular in almost every domain of life. Online shopping, e-commerce and transactions are increasing every day. Fraudulent activities have also increased in the online payment system. Payment card fraud has become a serious problem in the world. Companies and institutions lose huge amounts annually due to fraud and the fraudsters continuously seek new ways to commit illegal actions. The good news is that fraud tends to be perpetrated to certain patterns and that it is possible to detect such patterns, and hence fraud. In this paper, we present a way to detect fraudulent transactions by a neural network model. Artificial neural network (ANN), when trained properly can work like a human brain. They learn by example, like people and are known as a very good classifier. Among the main characteristics of credit card traffic are the great imbalance between proper and fraudulent operations. At the same time, public data are hardly available for confidentiality issues. To deal with the imbalanced dataset we use resampling techniques. To ensure proper model construction we made a pattern recognition network trained with a scaled conjugate gradient backpropagation algorithm using the Neural network toolbox in Matlab.