Open Internet by MindsNet

Improving GNN Model Performance for Fraud Detection

A research team is struggling to achieve high performance with their Graph Neural Network (GNN) model for fraud detection, despite using a well-known dataset and implementing necessary feature engineering. The model's metrics, such as AUC and precision, are lower than state-of-the-art (SOTA) models. The team is seeking help to identify potential issues with their approach.

Computing & Technology, Computer Science, Machine Learning