If you look again at the preceding image, you can see that the repository neo4j.github.com is shared between two people, while the parents repository would be recommended by only one person. This information can be used to rank the recommendations. The corresponding Cypher query would be as follows:
MATCH (user:User {login: "boggle"})-[:CONTRIBUTED_TO]->(common_repository:Repository)<-[:CONTRIBUTED_TO]-(other_user:User)-[:CONTRIBUTED_TO]->(recommendation:Repository)
WHERE user <> other_user
WITH recommendation, COUNT(other_user) as reco_importance
RETURN recommendation
ORDER BY reco_importance DESC
LIMIT 5
The new WITH clause is introduced to perform the aggregation: for each possible recommended repositories, we count how many users would recommend it.
This is the first way of using user data to provide accurate recommendations. Another way is, when possible, to take into account using social relationships, as we will see now.
