Practice monopartite projections

Degree centrality on movies

Run degree centrality on the 'movies-only' graph to find movies with the most connections to other movies (through shared actors):

CALL gds.degree.stream('movies-only') // (1)
YIELD nodeId, score // (2)
RETURN gds.util.asNode(nodeId).title AS movie, score AS connections // (3)
ORDER BY score DESC // (4)
LIMIT 10 // (5)

The top result is likely "True Romance"—an important film, yes, but is it truly the most important?

Remember what degree centrality measures: the number of connections each node has. In our Movie-to-Movie network, connections represent shared actors. "True Romance" ranks highly because it shares cast members with many other movies.

Even though "True Romance" is culturally important and critically acclaimed, that is not what PageRank measures on this projection. It means the film’s cast appeared in many other films in our database. Looking at the cast explains this:

The film features several highly-connected actors, creating many Movie-to-Movie connections.

The key takeaway here is that 'importance' or 'centrality' only matter in terms of the relationships analyzed. In this case, we have not returned the most important or central movies in terms of user ratings, or cultural influence.

PageRank on movies

PageRank also measures importance, but weights each node’s score based on the importance of its neighbours. Run PageRank on the same graph:

CALL gds.pageRank.stream('movies-only') // (1)
YIELD nodeId, score // (2)
RETURN gds.util.asNode(nodeId).title AS movie, score AS importance // (3)
ORDER BY score DESC // (4)
LIMIT 10 // (5)

The results differ from degree centrality because PageRank considers not just how many connections a movie has, but how important those connected movies are.

Why PageRank works here: This is a true monopartite graph. Movies connect to other movies, so PageRank’s "importance flow" can circulate through the network. Compare this to the previous lesson, where PageRank on a bipartite Actor→Movie structure produced meaningless results because importance flowed into Movies and got trapped.

However, note that PageRank on this particular graph produces essentially the same output as degree centrality. This happens because the Movie-to-Movie connections are relatively uniform in weight—movies connect through shared actors, and those actor-based connections don’t carry varying "importance" the way they might in other networks. In a network where some connections are inherently more valuable than others (like web pages linking to authoritative sources), PageRank would diverge more significantly from degree centrality.

Compare algorithms on user network

You have another graph in memory: 'user-network'. Run PageRank to find the most important users:

CALL gds.pageRank.stream( // (1)
  '?????' // (2)
)
YIELD nodeId, score // (3)
RETURN gds.util.asNode(nodeId).name AS user, score AS importance // (4)
ORDER BY score DESC // (5)
LIMIT 10 // (6)

Now run degree centrality on the same graph to compare:

CALL gds.degree.stream( // (1)
  '?????' // (2)
)
YIELD nodeId, score // (3)
RETURN gds.util.asNode(nodeId).name AS user, score AS connections // (4)
ORDER BY score DESC // (5)
LIMIT 10 // (6)

Comparing the results:

A user like "Angela Garcia" might have the highest degree (rated movies that many others also rated), while "Zachary Carey" might have the highest PageRank (connected to other highly-connected users).

Both are valid measures of "importance"—they just measure different things.