Repository files navigation

🚀Version 2.9.0 out now!Read the release notes here..

skpro is a library for supervised probabilistic prediction in python.It providesscikit-learn-like,scikit-base compatible interfaces to:

• tabularsupervised regressors for probabilistic prediction - interval, quantile and distribution predictions
• tabularprobabilistic time-to-event and survival prediction - instance-individual survival distributions
• metrics to evaluate probabilistic predictions, e.g., pinball loss, empirical coverage, CRPS, survival losses
• reductions to turnscikit-learn regressors into probabilisticskpro regressors, such as bootstrap or conformal
• buildingpipelines and composite models, including tuning via probabilistic performance metrics
• symbolicprobability distributions with value domain ofpandas.DataFrame-s andpandas-like interface

Overview
Open Source
Tutorials
Community
CI/CD
Code
Downloads
Citation

Questions and feedback are extremely welcome!We strongly believe in the value of sharing help publicly, as it allows a wider audience to benefit from it.

skpro is maintained by thesktime community, we use the same social channels.

Our objective is to enhance the interoperability and usability of the AI model ecosystem:

• skpro is compatible withscikit-learn andsktime, e.g., ansktime proba forecaster canbe built with anskpro proba regressor which in ansklearn regressor with proba mode added byskpro

• skpro provides a mini-package management framework for first-party implementations,and for interfacing popular second- and third-party components,such ascyclic-boosting,MAPIE, orngboost packages.

skpro curates libraries of components of the following types:

To installskpro, usepip:

pip install skpro

or, with maximum dependencies,

pip install skpro[all_extras]

Releases are available as source packages and binary wheels. You can see all available wheelshere.

fromsklearn.datasetsimportload_diabetesfromsklearn.ensembleimportRandomForestRegressorfromsklearn.linear_modelimportLinearRegressionfromsklearn.model_selectionimporttrain_test_splitfromskpro.regression.residualimportResidualDouble# step 1: data specificationX,y=load_diabetes(return_X_y=True,as_frame=True)X_train,X_new,y_train,_=train_test_split(X,y)# step 2: specifying the regressor - any compatible regressor is valid!# example - "squaring residuals" regressor# random forest for mean prediction# linear regression for variance predictionreg_mean=RandomForestRegressor()reg_resid=LinearRegression()reg_proba=ResidualDouble(reg_mean,reg_resid)# step 3: fitting the model to training datareg_proba.fit(X_train,y_train)# step 4: predicting labels on new data# probabilistic prediction modes - pick any or multiple# full distribution predictiony_pred_proba=reg_proba.predict_proba(X_new)# interval predictiony_pred_interval=reg_proba.predict_interval(X_new,coverage=0.9)# quantile predictiony_pred_quantiles=reg_proba.predict_quantiles(X_new,alpha=[0.05,0.5,0.95])# variance predictiony_pred_var=reg_proba.predict_var(X_new)# mean prediction is same as "classical" sklearn predict, also availabley_pred_mean=reg_proba.predict(X_new)

# step 5: specifying evaluation metricfromskpro.metricsimportCRPSmetric=CRPS()# continuous rank probability score - any skpro metric works!# step 6: evaluat metric, compare predictions to actualsmetric(y_test,y_pred_proba)>>>32.19

There are many ways to get involved with development ofskpro, which isdeveloped by thesktime community.We follow theall-contributorsspecification: all kinds of contributions are welcome - not just code.

To citeskpro in a scientific publication, seecitations.