A library of tools for easier evaluation of ML models.
Project description
MLLytics
Installation instructions
pip install MLLytics
or
python setup.py install
or
conda env create -f environment.yml
Update pypi instructions (for me)
Creates the package
python setup.py sdist bdist_wheel
Upload package
twine upload --repository pypi *version_files*
Future
Improvements and cleanup
- Allow figure size and font sizes to be passed into plotting functions
- Comment all functions and classes
- Add type hinting to all functions and classes (https://mypy.readthedocs.io/en/latest/cheat_sheet_py3.html)
- Example guides for each function in jupyter notebooks
MultiClassMetrics should inherit from ClassMetrics- REGRESSION
Cosmetic
- Fix size of confusion matrix
- Check works with matplotlib 3
- Tidy up legends and annotation text on plots
- Joy plots
- Brier score for calibration plot
- Tidy up cross validation and plots (also repeated cross-validation)
- Acc-thresholds graph
Big push
- Scoring functions
- MultiClassMetrics class to inherit from ClassMetrics and share common functions
- More output stats in overviews
- Update reliability plot https://machinelearningmastery.com/calibrated-classification-model-in-scikit-learn/
- Tests
- Switch from my metrics to sklearn metrics where it makes sense? aka
fpr, tpr, thresholds = roc_curve(y[test], probas_[:, 1])
and more general macro/micro average metrics from: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.recall_score.html#sklearn.metrics.recall_score - Additional metrics (sensitivity, specificity, precision, negative predictive value, FPR, FNR, false discovery rate, accuracy, F1 score
Contributing Authors
- Scott Clay
- David Sullivan
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
MLLytics-0.1.14.tar.gz
(11.3 kB
view hashes)
Built Distribution
Close
Hashes for MLLytics-0.1.14-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 86a6ebbbbc65320dcb5e1b804a1b23a717ff82544c070d41d0c2c4c53c107e12 |
|
MD5 | 86381b7006ab717c7985c8274d1c9461 |
|
BLAKE2b-256 | ffce9b99dcbd963bfc2481d0744b39c17c1ecf8769604683b0b090f72fd5379f |