pyroc is a package for analyzing receiver operator characteristic (ROC) curves. It includes the ability to statistically compare the area under the ROC (AUROC) for two or more classifiers.
Install:
pip install pyroc
Use:
import pyroc import numpy as np
pred = np.random.rand(100) target = np.round(pred) # flip 10% of labels target[0:10] = 1 - target[0:10] W = pyroc.auroc(target, pred)
# second prediction pred2 = pred pred2[10:20] = 1 - pred2[10:20] auroc, ci = pyroc.auroc_ci(target, [pred, pred2]) print(auroc) print(ci)
A usage.ipynb notebook is provided demonstrating common usage of the package (requires Jupyter: pip install jupyter).
Documentation is available on readthedocs. An executable demonstration of the package is available on GitHub as a Jupyter Notebook.
To install the package with pip, run:
pip install pyroc
To install this package with conda, run:
conda install -c conda-forge pyroc