Triplets similarities how to?

[steve5]

I have a lot of triplets (node, relation, node or circle, relation, circle) with different attributes for each user node (it can contain user_id, location, role for some nodes and others can be without location but with their marital status, use_car etc.) and data node.

Data node can contain location, size, origin and sometimes it will contain only the location. I have a relation between these nodes - that has some attributes like folder_name, approved/rejected etc.

Given a new triplet, how can I find the most similar triplets by their attributes (user node + relation + data node)

Is there any functionality to do this? I will be happy to get a direction to check or minimal example.

[dhruv_sharma882]

Hi Steve,

1- There is another discussion on the community forum similar to your question and has a few suggestions there. It can be found on->
How to find similarity between two graphs in Neo4j?.
This might provide you some options to try out and see what best works.

2- GDS (Graph data science) library of Neo4j also provides out of box support for some similarity algorithms that could used, which one best fits is subjected to the exact use case you want to achieve.

Neo4j Graph Database Platform

Node Similarity - Neo4j Graph Data Science

This section describes the Node Similarity algorithm in the Neo4j Graph Data Science library.

3- Apoc also provides graph comparison algorithms like fingerprinting, around which a solution can be created.

Neo4j Graph Database Platform

Fingerprinting - APOC Documentation

This section describes procedures that can be used to create a hash over nodes, relationships or the whole graph.

What would exactly fit for your use case is subjective to the problem, the solution looking for, the neo4j db performance, etc. I would recommend you to go through all these option and take the best 2-3 options and try them over a POC to see what best solves your problem and is also efficient with your current neo4j configurations.

Regards,
D