This archive contains code to reproduce the main results described in the article "NeuralSympCheck: A Symptom Checking and Disease Diagnostic Neural Model with Logic Regularization".

To reproduce these results, follow the steps described below.

Create an isolated environment with venv or conda. Install the dependencies listed in the src/requirements.txt

Download datasets from https://drive.google.com/drive/folders/19Jv_4wwC6LM485hDf8O5uhYb8QbAZOms?usp=sharing and put its to the folder data/05_model_input

Activate the environment and run this command at the root of the repository: kedro run --pipeline symptom_checker --params device:device,ds_name:ds_name,mode:test

Where:

• device - GPU idx or 'cpu'
• ds_name:
  • mz - for MuZhi dataset
  • dxy - for Dxy dataset
  • symcat_200 - for SymCat dataset with 200 deseases
  • symcat_300 - for SymCat dataset with 300 deseases
  • symcat_400 - for SymCat dataset with 400 deseases

The following results will be saved after the Pipeline is completed:

• model in the folder data/06_models
• metrics in the folder data/08_reporting