cancel
Showing results for 
Search instead for 
Did you mean: 

Head's Up! Site migration is underway. Expect disruption to service on Thursday, Feb. 9!

Neo4j v5.3 causing issues to Louvain algorithm?

aaronx-h
Node

I recently updated neo4j from v4.4.14 to v5.3 and came across an issue with the Louvain algorithm returning incorrect Louvain scores when comparing my manual algorithm with bloom's inbuilt algorithm (The communities generated by my code are different to the communities generated by blooms inbuilt algorithm). This used to work perfectly back in v4.4.14 where Bloom's Louvain score would match my manually generated Louvain scores. Just wanted to confirm if theres any problems with my graph projection and running of the algorithm here.

 

 

MATCH p=(n:Label)-[r:MST_BASIC]->(m:Label) 
WHERE r.date = '2021-02-04' 
SET n:Louvain 
SET m:Louvain

CALL gds.graph.project(
    'louvain',
    'Louvain',
    {MST_BASIC:
        {orientation:'UNDIRECTED',
        aggregation: 'NONE',
        properties:'distance'}
    }
)

CALL gds.louvain.write(
    'louvain', 
    {relationshipWeightProperty:'distance',
    writeProperty:'Louvain_Score'})
YIELD preProcessingMillis, computeMillis, writeMillis, communityCount, modularity
RETURN preProcessingMillis, computeMillis, writeMillis, communityCount, modularity;

 

 

The Louvain label just refers to nodes i've marked on a specific date that were a part of a Minimum spanning tree graph where MST_BASIC is the relationship type connecting the nodes. The relationship has a property of distance which is a result of the MST.

Any help would be appreciated!

 

0 REPLIES 0
Nodes 2022
Nodes
NODES 2022, Neo4j Online Education Summit

All the sessions of the conference are now available online