Splitting the graph and computing the score for each edge

In order to make predictions using a link prediction score, we need to follow these steps:

1. Define the state of the graph at a given time, t₁
2. At time t₁, compute the link prediction score for each pair of nodes in the graph
3. Compare these predictions to the links created between t₁ and t₂
4. Determine a score threshold such that pairs with a score above this threshold are more likely to appear at time t₂ > t₁

In this example, the state of the graph at t₁ is already provided through the KNOWS_T1 relationships.

We can now compute the link prediction score for each pair of nodes in the graph, considering only the knowledge at time t₁ by using the newly created relationship. Here, we choose to use the Adamic-Adar score:

MATCH (u) MATCH (v) WHERE u <> v
RETURN  u.id as u_id, 
        v.id as v_id,  
        gds.alpha.linkprediction.adamicAdar(u, v, {
            relationshipQuery: "KNOWS_T1",
            direction: ...