A powerful application of graph data science is graph completion, using supervised machine learning. The structure of your connected data serves as your labeled data, and your model is trained to fill in the blanks: missing node labels and missing relationships.
One common use case for graph completion is entity resolution (ER). ER allows businesses to consolidate customer profiles across multiple data streams to create unique, valuable entity profiles.
Graph-based approaches to entity resolution allow you to use not only the traditional identifiers of an entity – such as names, addresses, and other personal identifiable information – but also actions and behavior to literally “connect the dots” between entities.
In this demo, I’ll show how to use Neo4j’s Graph Data Science (GDS) library to rapidly develop supervised ML pipelines. We’ll use ER as a case study and walk through an example to demonstrate how it could be applied to your own data:
Quick overview of Entity Resolution (ER) and ER in Graph
Dive into an example based on real-world data where we will use GDS Link Prediction Pipelines to train an entity linkage model and predict new entity links in the graph
Go over quick procedures for generating resolved entity
Query resolved entities out of the graph
Do you want to replicate the demo yourself? I have posted the source code on Github. https://github.com/zach-blumenfeld/demo-lp-for-entity-resolution