Hi, Time complexity Similarity algorithms has time complexity O(n3) and space complexity O(n2). neo4j is a graph database and all the nodes are connected other nodes with respect to their relationship. Basically I wanted to build a content based movies recommender systems. content based RS is only dependent on user personalised content and his history. I had created a movie graph database in neo4j and i had two possibilities one is recommend 1st, 2nd connectioned movie nodes that the user watched movies or another solution is to use any node similarity like jaccard. I'm little bit confused why we need node similarity in the first place when all the nodes are connected to each other, i mean we can use some constraints to return the closest node data right?
An appropriate similarity metric would depend on your data model and use case. For a bipartite graph, the Jaccard measure (or something similar) seems to make a lot of sense to me. You are measuring pair-wise similarity based on the number of similar items they have in common compared to the total number of items the both are related to. As such, the similarity score will be a number between 0 and 1 (inclusive of both). You could extend this to make it more complicated, but still applicable, if you have a more complicated definition of "items in common" versus "related items."