Collaborative filtering algorithms are prone to mainstream-taste bias

Source code for the RecSys 2023 paper Collaborative filtering algorithms are prone to mainstream-taste bias by Pantelis P. Analytis and Philipp Hager.

Installation

1. Install dependencies using conda: conda env create -f environment.yaml
2. Activate environment: conda activate mainstream-taste-bias
3. Run experiments as described below.

Run experiments

1. Evaluate all models and store user-level evaluation results:

python main.py -m \
    data=faces,jester,movielens \
    model=ease,funk,knn-item-item,knn-user-user,nmf

2. Plot results:

python plot-figure-1.py

3. Evaluate how predictive user features are of user-level model performance:

python r2_analysis.py

Reference

@inproceedings{Analytis2023MainstreamTasteBias,
  author = {Pantelis P. Analytis and Philipp Hager},
  title = {Collaborative filtering algorithms are prone to mainstream-taste bias},
  booktitle = {Proceedings of the 17th ACM Conference on Recommender Systems (RecSys`23)},
  organization = {ACM},
  year = {2023},
}

License

This project uses the MIT license.