A custom taxonomy-driven GraphSAGE framework for population-scale POI co-visitation prediction. Features multi-modal node/edge embeddings integrating NAICS business classification taxonomy, deep edge regression pipelines, and advanced loss function optimization for extreme class imbalance.
geometric-deep-learning mobility-modeling graph-neural-networks applied-ai pytorch-geometric gnn edge-regression taxonomy-driven-graphsage naics-embeddings
-
Updated
Jun 2, 2026 - Python