CODEWITHSUNDEEP
elasticsearchrecommendation-engine
Chirag Jain
Chirag Jain

Reputation: 2418

Elasticsearch: suggest users based on likes

I'm trying to model user suggestion system in elasticsearch which takes into account users likes and profile.

I have the user structure like this:

user: { 
    id: 232344,
    location: 'New York',
    likes: [4545, 3434, 343]
}

I want to suggest users based on below three concepts:

1) users who liked me.

2) users like the one i liked. (similar location, etc)

3) mutual likes. (user a like user x. user b likes user x. suggest user a to b and vice versa.)

I have read about More like this query in elasticsearch but i'm not sure if it can take into account all these scenarios.

How can i model these things in elasticsearch queries or should i consider using a graph database like neo4j?

Upvotes: 0

Views: 611

Answers (1)

moliware
moliware

Reputation: 10288

1) Users who like me. Suppose that my id is 1

{ 
    "query": {
        "term" : { "likes" : 1 }
    }
}

2) Check more like this api. Given a document id and a set of fields it responds with similar documents to the given one. There are lots of options so I recommend to read the documentation and see if it fits your use case.

3) mutual likes: Suppose I'm a and I liked these users [1, 3, 5]. This query will return b if b also liked one of those ids.

{ 
    "query": {
        "term" : { "likes" : [1, 3, 5] }
    }
}

Upvotes: 1

Related Questions