To run all methods on the example dataset, Iris:
$ python run.py 1 True iris
Some dependent libraries for baselines:
$ pip install pyarc # For CBA
$ conda install orange3 # For CN2
IDS is adapted from here. Their code is too old to run and has bugs. I fixed them.
Project organization:
- Data is preprocessed in data-multi.ipynb
- Metrics are generated in table.ipynb. However you need Sacred to manage the experimental data.
In case you wonder, this repo adopts sacred and incense to manage experiment results. However, you don't need to follow the same routine. Without Sacred, results will be redirected to stdout.
$ pip install dnspython incense sacred
# a private file: mongodburi.py
mongo_uri = 'mongodb+srv://xxx'
db_name = 'yyy'
This repo also uses Python type checking feature mainly for annotation and pytest for testing.
pip install pytest
$ pytest
Datasets can be downloaded from the following list.