This repository is dealing with reproducing the paper Data-Driven Methods for Balancing Fairness and Efficiency in Ride-Pooling. This implementation is based on the official code.
To setup the environment, install anaconda and run the following command while being at the root directory of this repository:
conda env create -f fact.yml
If you want the environment with GPU support, run the following command instead:
conda env create -f fact_gpu.yml
Create the subdirectory 'data' in the root directory of this repository. In order to provide the actual samples, the following datasets have to be downloaded.
The yellow taxi trip records dataset is used for the experiments. The preprocessed dataset can be requested from the authors of the original paper and has to be placed in 'data/ny'.
Download the dataset Demographics by Neighborhood Tabulation Area dataset and place it in 'data/demo'.
Download the dataset Neighborhood Tabulation Areas (NTAs) dataset and place it in 'preprocessing'.
Our pretrained models are in the models subdirectory of this repository.
The jupyter notebook run_experiments.ipynb in the folder 'src' provides an interface for preprocessing, training and evaluation. If pre-trained models are available they will be used to generate the results. The notebook can be opened by activating the environment:
conda activate fact
and executing:
jupyter notebook
This opens a tab in the browser where the notebook file can be opened. By running all cells, our results can be obtained.
The following Figures show our reproduced results corresponding to Figures 1, 2 and 3 in the original paper.
