scoreTree is a Python package for the method introduced in the paper titled 'Building Trees for Probabilistic Prediction via Scoring Rules.'

This code requires Python (version 3.6 or later) and pip.

Examples under examples/ directory replicate Figure 1 and Figure 4.

We recommend creating a Python virtual environment within the working directory of scoreTree. If a virtual environment is created, scoreTree's required packages are installed and isolated from those installed a priori. Creating a virtual environment will also prevent having conflicting packages on a user's machine. You may need to install the virtual environment on your system (if a user's system does not have it), for example, with 'apt install python3.9-venv'

1)Extract the zipped file.

2)From the command line, go to the directory of the source code.

3)Use the following command to create a virtual environment:

python3 -m venv venv/
source venv/bin/activate

We note that creating a virtual environment is not a required step. However, we tested this procedure to prevent any conflict, and the code runs smoothly.

To install the package:

1)Go to the directory of the source code (if a user has not done so yet).

2)Use the following command to install the required packages:

pip install -r requirements.txt

3)From the command line, use the following command to install the package:

pip install -e .

Once installed, a user should see build/ directory created.

To replicate Figure~1:

1)Go to the examples/ directory.

2)Execute the followings from the command line:

python3 Figure1.py

Running this script should not take more than 60 sec. See the figures (png files) saved under examples/ directory.

Instructions are provided to replicate Figure~4.

Execute the following from the command line:

python3 Figure4a.py

Running this script takes about 2.5hrs on a personal Mac laptop. Once completed, Figure4a.png is saved under examples/ directory. Moreover, Table~3 (easy dataset) is printed to the screen.

Execute the following from the command line:

python3 Figure4b.py

Running this script takes about 2.5hrs on a personal Mac laptop. Once completed, Figure4b.png is saved under examples/ directory. Moreover, Table~3 (hard dataset) is printed to the screen. Finally, Figure 2 Figure2.png is saved under examples/ directory.

Type deactivate from the command line to deactivate the virtual environment if created.

Type pip uninstall scoreTree from the command line to uninstall the package.