Title

AWS re:Invent 2022 - Graph feature engineering with Neo4j and Amazon SageMaker (PRT022)

Summary

• The session covered graph feature engineering using Neo4j and Amazon SageMaker.
• The speakers discussed the integration of Neo4j with AWS services and provided a demonstration.
• Neo4j's components include a graph database, a data science component for machine learning, and Neo4j Bloom for visualization.
• The architecture involves pulling data from an Amazon S3 bucket into SageMaker Studio, processing it with Neo4j's graph data science, and then using SageMaker Autopilot for automated training and inference.
• Neo4j is available on the AWS Marketplace for easy deployment within a user's VPC.
• The demo showcased deploying Neo4j Enterprise Edition through AWS Marketplace, loading data from the SEC's EDGAR system, and using the FastRP algorithm to create vectors for machine learning.
• The session concluded with next steps, including trying Neo4j on AWS Marketplace and visiting the Neo4j and Amazon landing page.

Insights

• Neo4j's integration with AWS services, such as S3 and SageMaker, allows for seamless graph feature engineering, which can enhance machine learning models by leveraging the relationships within data.
• The use of Neo4j Bloom for visualization can help users better understand complex graph relationships and the results of their machine learning models.
• The FastRP algorithm mentioned in the demo is a graph embedding technique that transforms graph data into a format suitable for traditional machine learning methods, which can be particularly useful for feature engineering.
• The ability to deploy Neo4j from the AWS Marketplace with a few clicks simplifies the setup process and encourages experimentation and adoption.
• The example provided in the demo, predicting portfolio churn from SEC filings, illustrates the practical application of graph feature engineering in the finance sector.
• The session highlighted the importance of accessible and reproducible examples, as evidenced by the use of a public GitHub repository and the provision of slides on social media.
• The mention of upcoming connectors and services at the next re:Invent suggests ongoing development and enhancements in the Neo4j and AWS partnership, which could lead to more advanced graph analytics capabilities in the future.