biolab / orange3-conformal

Conformal Prediction library for Orange3

Orange3 Conformal Prediction

Conformal Prediction is an add-on for Orange3 data mining software package. It provides an extensive toolset for conformal prediction.

Installation

To install the add-on, run

python setup.py install

To register this add-on with Orange, but keep the code in the development directory (do not copy it to Python's site-packages directory), run

python setup.py develop

Usage

The library in the add-on can be used in Python scripts. The add-on does not provide any GUI widgets.

The example below evaluates an inductive conformal predictor at 0.1 significance level on the Iris dataset (spliting it into a training and testing set in ratio 2:1). The nonconformity scores used by the conformal predictor are based on the probabilities returned by a Naive Bayes classifier.

import Orange
import orangecontrib.conformal as cp

tab = Orange.data.Table('iris')
nc = cp.nonconformity.InverseProbability(Orange.classification.NaiveBayesLearner())
ic = cp.classification.InductiveClassifier(nc)
r = cp.evaluation.run(ic, 0.1, cp.evaluation.RandomSampler(tab, 2, 1))
print(r.accuracy())

Documentation

Please see doc/Orange-ConformalPrediction.pdf. Documentation in other formats can also be built using Sphinx from the doc directory.

Online documentation is available at https://orange3-conformal.readthedocs.io.