how can I use algo to do community detection for subgrap which select by match cypher ?

[heyichang]

CALL algo.louvain( 'MATCH (n:Account) RETURN id(n) as id', 'match (n1:Account{nodeID:"BDB8106C5276AE4AB67EF8FE9F7E4282"})-[r1:follow]->(n2:Account)-[r2:follow]->(n3:Account) RETURN id(n1) as source, id(n3) as target, count(*) as weight', {graph:'cypher', iterations:5, write: true});

but this can't work ,

loadMillis	computeMillis	writeMillis	nodes	iterations	communityCount
171	125	85	29889	0	29889

I have 29889 nodes and also have community count,it didn't select graph and do in this

[heyichang]

And then ,I change my cypher

CALL algo.louvain(
'MATCH (n:Account{nodeID:"BDB8106C5276AE4AB67EF8FE9F7E4282"}) RETURN id(n) as id union match (n1:Account{nodeID:"BDB8106C5276AE4AB67EF8FE9F7E4282"})-[r1:follow]->(n2:Account) return id(n2) as id union match (n1:Account{nodeID:"BDB8106C5276AE4AB67EF8FE9F7E4282"})-[r1:follow]-(n2:Account)-[r2:follow]-(n3:Account) return id(n3) as id ',
'match (n1:Account)-[r1:follow]->(n2:Account)-[r2:follow]->(n3:Account)
RETURN id(n1) as source, id(n3) as target',
{graph:'cypher', iterations:100, write: true});

And it is also can't get the good effect

[heyichang]

the consequence

loadMillis  computeMillis   writeMillis nodes   iterations  communityCount
2748    14020   257 19705   3   19698

[mknblch]

Hi. In my tests, whenever the community count is equal or almost equal to the count of nodes there is something wrong with the relationship import. Either due to a wrong query or a bug. The algorithm sets the node-id as community-id at statup but if important relationships are missing the algo never touches those nodes again. Which, in turn, ends up in a wrong community count (and short computation duration).

On the first look it seems your relationship query is too specific. Could you possibly create a testgraph which reproduces the problem? You could also try to relabel your subgraph and call the algorithm with the new label and rel-name.