Triadic closure - find triangles with edge properties

[georg_kf_heiler]

How can I find all triangles in Neo4j including the edges (and edge properties)

CALL gds.graph.create('g-movies',
    ['Person', 'Movie'],
    {PARTICIPATE_IN_CRIME:{
        type:"ACTED_IN",
        orientation:"UNDIRECTED"
    }
});
        
CALL gds.triangleCount.write('g-movies', {
  writeProperty: 'triangles'
})
YIELD globalTriangleCount, nodeCount;

CALL gds.alpha.triangles('g-movies')
YIELD nodeA, nodeB, nodeC
RETURN
  gds.util.asNode(nodeA) AS nodeA,
  gds.util.asNode(nodeB) AS nodeB,
  gds.util.asNode(nodeC) AS nodeC

LIMIT 25
;

already returns the triangles (on this movie's dummy graph it is 0) (= I was not able to create a person-to-person projection of the graph quickly). But in principle, this returns the 3 nodes forming the triangle.

How can I get the edge properties for these edges as an additional column - i.e. assuming ACTED_IN would hold a property like roles (or date) I would want to retrieve these properties as an additional column i.e. AB_roles, AC_roles, BC_roles. How can I formulate this in neo4j/apoc/cypher?

[david_rosenblum]

If we want to look at just Persons as a mono partite graph - you can do a cypher projection

CALL gds.graph.create.cypher('g-movies5',
'match (n:Person)return id(n) as id','match (s:Person)--(:Movie)--(t:Person) return id(s) as source, id(t) as target '
);
We then get a lot of triangles.
Did you then want something like all the Persons in Triangle Group 1?

[georg_kf_heiler]

david.rosenblum:

CALL gds.graph.create.cypher('g-movies5',
'match (n:Person)return id(n) as id','match (s:Person)--(:Movie)--(t:Person) return id(s) as source, id(t) as target '
);

Great for sharing this!
After obtaining the triangles from the person to person projected data (I did this projection outside of neo4j so far) but with this I will perhaps do so in the future natively inside neo4j, I want to add additional columns /attriutes to the edges i.e. AB_attr1, AC_attr1, BC_attr1 to compare some attributes (such as timestamps) on the edges.

If I would do it outside neo4j I simply would join i.e. in pandas 3x the edges with the triangles obtained in neo4j and then obtain the desired columns/attribtes on the 3 edges.

But is there a more native way to do so directly?