How does the BIS Innovation Hub’s Project Aurora utilize synthetic data in its analysis of fraudulent payments?

The BIS Innovation Hub’s Project Aurora uses synthetic data to assess the effectiveness of various models in detecting fraudulent payments. It compares traditional models with machine learning techniques, including isolation forests and neural networks, and employs graph neural networks to analyze suspicious transaction networks.

Project Aurora leverages synthetic data to evaluate different models aimed at identifying fraudulent payments. The project contrasts traditional detection methods with advanced machine learning techniques, such as isolation forests and neural networks. Additionally, it utilizes graph neural networks to uncover suspicious transaction networks and employs autoencoder models to detect anomalies by analyzing input and output data, effectively recognizing patterns indicative of fraud.