find similar weighted nodes in neo4j

Question

• Nike Sells(weight .9) Shoes
• Nike Sells(weight .5) Apparel
• Reebok Sells(weight .9) Shoes ...

We are talking about 10k brands and around 100k relations.

Now given a Brand like Puma I need to find brands which are similar to Puma based on what they sell.

I read about general similarity search, but didn't get how we can include weights in the query.

Answer 1

MATCH (b:Brand {name: 'Puma'})
WITH b
MATCH x = (b) - [:SELLS] -> (:Item) <- [:SELLS] - (ob:Brand)
RETURN b, ob, ABS(SUM(RELATIONSHIPS(x)[0]['weight'] - RELATIONSHIPS(x)[1]['weight']))

You can change how the difference in weights is scored, but that's how you can at least collect all of the paths to other brands to get the weights to compare.