Graph analytics

Build explainable graph views and use the same temporal provenance in a graph database or a PyTorch Geometric experiment.

Time	Extras	Output
30–45 min	base; Neo4j optional	graph tables, Cypher, fingerprint
30–45 min	graph optional	temporal features and evaluation

The temporal graph notebook includes the optional !pip install neo4j cell, a local Neo4j startup command, connectivity check, and a small Cypher query. The offline graph export and Polars investigation remain the default path. The PyTorch Geometric notebook uses the existing graph extra; no external service is required.

Tutorials

• Investigate fraud with temporal graph exports

• Build and evaluate temporal graph features

Related: graph and benchmark workflows.

Next path: Streaming and Kafka reliability.