Reproducing the results of the kickscore paper

This repository contains the code to reproduce the results described in the paper

Lucas Maystre, Victor Kristof, Matthias Grossglauser, Pairwise Comparisons with Flexible Time-dynamics, KDD 2019.

Instructions

Start by cloning the Git repository:

git clone https://github.com/lucasmaystre/kickscore-kdd19.git

Next, download the data from Zenodo. Extract the contents of the archive inside the top-level directory of the repository, by using

tar xvf kickscore-kdd-20190725.tar.gz

Set up an environment variable pointing to the data folder:

export KSEVAL_DATASETS=path/to/data

You might want to put this line in your ~/.bashrc to avoid typing it every time you open a terminal. Then, install the required Python libraries:

pip install -r requirements.txt

Now you can start a notebook server as follows:

cd notebooks
jupyter notebook