Online Fraud Detection using Deep Learning Techniques

Fraud has no permanent patterns. They are constantly changing their behavior; therefore, we need to use unsupervised reading. Fraudsters learn new technologies that allow them to commit fraud through online transactions. Fraudsters consider the common behavior of consumers, and fraudulent methods are changing rapidly. Therefore, fraudulent schemes need to detect online transactions through unsupervised learning, as some scammers commit fraud as well as using internet users and switch to other strategies. This paper aims to 1) focus on fraud cases that are not available based on previous history or supervised learning, 2) to create a deep and restricted Auto-encoder model with Boltzmann (RBM) machine that can recreate standard transactions to detect irregularities from common patterns. The proposed in-depth instruction based on auto-encoder (AE) is an unregulated learning algorithm that works on previous applications by setting inputs equal to the results. RBM has two layers, input (visible) layer and hidden layer.