DmitriiDenisov / research_oversampling

Repository for research of oversampling technique with Gamma distribution

Oversampling research

How to run:

Run: python3 main.py

This will produce a file compare_temp/output_{k}_{theta}_success_{success}_seed_{seed_value}.xlsx

Execution time on server ~ 1 hour

Dependencies:

Python3.7.7

pip install -r requirements.txt

Structure of Repo:

main.py - script which runs all experiments and produces results into excel file

test.py - script for tests implemented with pytest library.

To run: pytest tests.py -vv

utils folder contains secondary functions which are used in main.py

tests folder contains functions only for tests which are used in test.py

experiments contains several jupyter notebooks only for visualization and experiential purposes

Authors:

Firuz Kamalov [e-mail]

Dmitry Denisov [e-mail]

Possible issues:

Problem: from imblearn.datasets import fetch_datasets does not work for me as inside it fails in importing from sklearn.utils.fixes import makedirs

Solution: Forcing pip install imbalanced-learn==0.5.0. The problem was observed on 0.4 version

Problem: urllib.error.URLError: <urlopen error [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed

Solution: https://stackoverflow.com/questions/50236117/scraping-ssl-certificate-verify-failed-error-for-http-en-wikipedia-org

Additional materials on SMOTE and ADDASYN

https://imbalanced-learn.readthedocs.io/en/stable/auto_examples/over-sampling/plot_comparison_over_sampling.html#sphx-glr-auto-examples-over-sampling-plot-comparison-over-sampling-py

https://youtu.be/FheTDyCwRdE

https://basegroup.ru/community/articles/imbalance-datasets