search

benfred / implicit

Fast Python Collaborative Filtering for Implicit Feedback Datasets
https://benfred.github.io/implicit/
MIT License
3.57k stars 612 forks source link

Recommend users to items #548

Open dragonlee97 opened 2 years ago

dragonlee97 commented 2 years ago

Hi, I know that it seems illogical, but in come other contexts, we would like to recommend user to a project/team/company. I am wondering is there a reversed build-in function of recommend() ? If not, what would be the most efficient the way to do this instead of reproducing the matrix with user factors and item factors and rank user by items?

thiagorizuti commented 2 years ago

I think you can transpose the data. Pass a user_item matrix to "fit", then pass a item_user matrix to "recommend" considering "userid" as "itemid".

elniakim commented 2 years ago

@dragonlee97 I don't think it's necessary to interpret the user literally in the recommendation system. So, in your case now, the company could be a "user" in a way, and the user you want to recommend can be viewed as an "item". Then, naturally, the same structure that recommends items to users, I wonder if we can go to a structure that recommends users to a company.