reranking 0.2.8

No project description provided

Reranking: fairness/personalization for recommendation and search

reranking provides algorithms to re-rank the items to any desired item attribute distribution.

This package can be used as a post-processing modular of recommendation system or search engine.

Inspired by paper Fairness-Aware Ranking in Search & Recommendation Systems with Application to LinkedIn Talent Search.

Concept of usage

Fairness: impose a same distribution to the ranking of each user

Taking "recommend candidates to recruiters" as the example (this is the case of LinkedIn Recruiter service stressed in the paper), the search engine ranking are re-ranked by the distribution of the protected attributes like gender and demographic parity in consideration of the fair display of the candidates to recruiters.

Personalization: impose distribution to each user based on their preference

For example, when we recommend products to users, the item preference distribution for each user (obtained by the purchase/view log) can be used to re-rank the item rankings calculated by recommendation systems.

Installation

$ pip install reranking

Examples

from reranking.algs import Reranking
r = Reranking(["f1", "f1", "f1", "f2", "f1", "f1", "f1", "f2"], {"f1": 0.5, "f2": 0.5})
r.re_rank(k_max=4) # we want "f1" and "f2" have equal proportion in top-4

The output is [0, 3, 1, 7] which is the re-ranked ranking of the top-4 items.

More examples can be found here.