Skip to content
This repository has been archived by the owner on May 1, 2023. It is now read-only.

Aarhus-Psychiatry-Research/psycop-model-training

Repository files navigation

Current priorities are on the Board 🎬.

psycop-model-training

python versions Code style: black [Tests][tests]

Prediction of type 2 diabetes among patients with visits to psychiatric hospital departments.

Using the package

This is a set of modules used for some of the projects' model training. You need project-specific code to use these modules. To get started with that, see template-model-training.

Installing to src

pip install --src ./src -e git+https://github.com/Aarhus-Psychiatry-Research/psycop-model-training#egg=psycop_model_training

Developing new evaluations

In general, model evaluations are added as their own file in

src > psycop_model_training > model_eval > base_artifacts > plots/tables

To make sure they run every time, also add them to base_artifact_generator.py.

However, when developing, it's much faster to develop on a synthetic dataset.

To do that:

Work locally

  1. Write your plot function in an appropriate file in the src > psycop_model_training > model_eval > base_artifacts > plots/tables directory.
  2. Test the plot on synthetic prediction data. Write a test in tests > model_evaluation > test_name_of_your_plot.py. Use the other visualization tests as a guide.

Work remotely

  1. When you're happy with the plot, test it on real data on Overtaci. To do this, go to Overtaci and replace the path in your test script with some real model predictions with metadata.