marinimau / fairrecsys

FairRecSys

Brief Description

FairRecSys is a post processing method in order to have less discriminative recommendations.

Requirements

• Python
• google ortools
• pandas
• numpy
• sklearn
• scipy

Usage

To run algorithm with metric \mu_1:

$ python alg_distMet_mu1.py \
altered_matrix_path \
userFile_path \
predictions_path \ 
topK \
epsilon

To run algorithm with metric \mu_2:

$ python alg_distMet_mu2.py \
altered_matrix_path \
userFile_path \
predictions_path \ 
topK \
epsilon

Explanation for Input parameters:
"altered_matrix_path": path to output altered matrix
"userFile_path": path to user file where sensitive attribute exists
"predictions_path": original recommendation matrix
"topK": top-k
"epsilon": desired level of fairness

Example:

$ python alg_distMet_mu1.py altered_fair_matrix ../data/ml-1m-users.csv ../data/ml-1m-pred-wrmf.csv 20 0.2

Questions

[e-mail]