actor_weightings.adoc

Movie recommendation based on User preferences and Actor weights

Given a movie database, with a structure like

(actor:Actor)-[:ACTS_IN { percent: y }]→(film:Movie) so that the system knows how long each actor is physically present on the screen, and to what % of the movie duration that corresponds.

Each user can have preferences for some actors (ie: grade them from "bad" to "awesome" or whatnot). To each of this preference the system is associated to a modifier (say from -100 to +100).

CREATE (m1:Movie{name:'m1'})
CREATE (m2:Movie{name:'m2'})
CREATE (m3:Movie{name:'m3'})
CREATE (a1:Actor{name:'a1'})
CREATE (a2:Actor{name:'a2'})
CREATE (a3:Actor{name:'a3'})
CREATE (u1:User{name:'u1'})
CREATE (u2:User{name:'u2'})
CREATE (u3:User{name:'u3'})
CREATE (a1)-[:ACTS_IN { percent: 10 }]->(m1)
CREATE (a1)-[:ACTS_IN { percent: 20 }]->(m2)
CREATE (a2)-[:ACTS_IN { percent: 20 }]->(m2)
CREATE (a3)-[:ACTS_IN { percent: 20 }]->(m1)
CREATE (u1)-[:VALUES { modifier: -20 }]->(a1)
CREATE (u1)-[:VALUES { modifier: 20 }]->(a2)
CREATE (u1)-[:VALUES { modifier: 100 }]->(a3)

What are the best movies for Users?

This example shows how to ask the database to return the first movies, ranked by the value given by each preferred actor modifier multiplied by its presence in % on the movie? (Giving 0 to those movies that have no actor in the user preferences.)

MATCH (u:User)-[v:VALUES]->(a:Actor)-[r:ACTS_IN]->(m:Movie)
WITH u, v.modifier*r.percent as score, a, m
RETURN u as user, SUM(score) as totalScore, collect(a) as actors_in_movie, m.name as movie
ORDER BY totalScore DESC

Full context at Stack Overflow