Adversarial Attention-Based Variational Graph Autoencoder for Fraud Detection in Online Financial Transaction Nur Alibasyah Wiriaatmadja (a*), Finny Oktariani (a,b)
a) Department of Computational Science, Bandung Institute of Technology
Jalan Ganesha 10, Bandung 40132, Indonesia
*nuralibasyah[at]gmail.com
b) Combinatorial Mathematics Research Group, Faculty of Mathematics and Natural Sciences, Bandung Institute of Technology
Jalan Ganesha 10, Bandung 40132, Indonesia
Abstract
The significant growth of online financial transactions also raises threats of fraud in transactions which can harm both companies and consumers. Several machine learning models for classification has been used for fraud detection and shows outstanding results. However, as the volume and variety of transaction data expands, it is difficult for traditional machine learning algorithms to detect fraud. This research aims to use the Adversarial Attention-Based Variational Graph Autoencoder (AAVGA) for fraud detection by modeling online transaction data as a graph and use the graph embedding for classification task. AAVGA is one of the most recent autoencoder-based graphs embedding method, which use graph attention network in the encoder to generate latent variable and use them in adversarial strategy to enhance the generalization performance of a graph embedding model. By its attention mechanisms, AAVGA can achieve promising accuracy results for unbalanced classification tasks, such as fraudulent transaction data. Furthermore, this research will try to show broad capabilities of graph embedding methods by modeling online transaction data in graph structure.
