Automated Fraud Detection in Online Transaction Using Representation Learning

Automated fraud detection in online transactions is a challenging task, as fraudsters are constantly developing new and sophisticated methods. Representation learning is a machine learning technique that can be used to learn patterns in data by transforming the data into a lower-dimensional representation that is more informative and easier to learn from. Two popular representation learning techniques are autoencoders and graph neural networks (GNNs).Auto encoders are a type of neural network that is trained to reconstruct its input. This means that an auto encoder learns to represent the input data in a compressed form, and can then be used to detect anomalies in the data. GNNs are a type of neural network that can be used to learn from data that is represented as graph. A graph is a data structure which in form of nodes and edges. This paper proposes a novel fraud detection system using auto encoders and GNNs. The proposed system is evaluated on a real-world dataset of online transactions, and is shown to achieve state-of-the-art results. We aim to capture complex relationships among accounts, transactions, and entities within the transaction network. The implementation of automated fraud detection using representation learning stands to enhance customer trust, reduce operational costs, and safeguard online transactions from the ever-evolving threat landscape.