Extracting features and labels

In order to create a dataset for a link prediction task, we need to do the following:

1. Compute the score for each pair of nodes in the graph, using the KNOWS_T1 relationships only.
2. Discard the pairs of nodes already linked to each other at t_1.
3. Extract the label for each of the remaining pairs; the label is True if a relationship between the two nodes exists at time t₂, otherwise False.
   

The following query performs these three operations:

MATCH (u)
MATCH (v) 
// take only one link from undirected graph
WHERE u.id < v.id  // exclude u = v
// exclude edges that were already there at T1:
AND NOT ( (u)-[:KNOWS_T1]-(v) )
// compute score
WITH u, v,  gds.alpha.linkprediction.adamicAdar(
                u, v, {
                        relationshipQuery: "KNOWS_T1", 
                        direction: "BOTH"
                }
            ) as score
RETURN  u.id as u_id, 
        v.id as v_id, 
        score, 
        // get the label: does the edge exist...