GraphFAS is a distributed system that generates interpretable structural features through non-parametric multi-hop subgraph extraction. It uses a modified Boruta algorithm to identify informative features, allowing direct integration with tabular models without requiring end-to-end GNN training.
HOW THIS AFFECTS YOU
●
builderYou can automate feature engineering for fraud detection while maintaining the interpretability of tabular models.