Scikit-survival

This release adds support for scikit-learn 0.24 and Python 3.9. scikit-survival now requires at least pandas 0.25 and scikit-learn 0.24. Moreover, if [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis) or [sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis.html#sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis) are fit with `loss='coxph'`, <span class="title-ref">predict\_cumulative\_hazard\_function</span> and <span class="title-ref">predict\_survival\_function</span> are now available. [sksurv.metrics.cumulative_dynamic_auc](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.metrics.cumulative_dynamic_auc.html) now supports evaluating time-dependent predictions, for instance for a [sksurv.ensemble.RandomSurvivalForest](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.RandomSurvivalForest.html) as illustrated in the [User Guide](https://scikit-survival.readthedocs.io/en/v0.15.0/user_guide/evaluating-survival-models.html#Using-Time-dependent-Risk-Scores).

Bug fixes

- Allow passing pandas data frames to all `fit` and `predict` methods (\148).
- Allow sparse matrices to be passed to [sksurv.ensemble.GradientBoostingSurvivalAnalysis.predict](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis.predict).
- Fix example in user guide using GridSearchCV to determine alphas for CoxnetSurvivalAnalysis (\186).

Enhancements

- Add score method to [sksurv.meta.Stacking](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.meta.Stacking.html), [sksurv.meta.EnsembleSelection](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.meta.EnsembleSelection.html), and [sksurv.meta.EnsembleSelectionRegressor](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.meta.EnsembleSelectionRegressor.html) (\#151).
- Add support for <span class="title-ref">predict\_cumulative\_hazard\_function</span> and <span class="title-ref">predict\_survival\_function</span> to [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis). and [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis.html#sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis) if model was fit with `loss='coxph'`.
- Add support for time-dependent predictions to [sksurv.metrics.cumulative_dynamic_auc](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.metrics.cumulative_dynamic_auc.html) See the [User Guide](https://scikit-survival.readthedocs.io/en/v0.15.0/user_guide/evaluating-survival-models.html#Using-Time-dependent-Risk-Scores) for an example (\134).

Backwards incompatible changes

- The score method of [sksurv.linear_model.IPCRidge](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.linear_model.IPCRidge.html), [sksurv.svm.FastSurvivalSVM](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.svm.FastSurvivalSVM.html), and [sksurv.svm.FastKernelSurvivalSVM](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.svm.FastKernelSurvivalSVM.html) (if `rank_ratio` is smaller than 1) now converts predictions on log(time) scale to risk scores prior to computing the concordance index.
- Support for cvxpy and cvxopt solver in [sksurv.svm.MinlipSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.svm.MinlipSurvivalAnalysis.html) and [sksurv.svm.HingeLossSurvivalSVM](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.svm.HingeLossSurvivalSVM.html) has been dropped. The default solver is now ECOS, which was used by cvxpy (the previous default) internally. Therefore, results should be identical.
- Dropped the `presort` argument from [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.tree.SurvivalTree.html) and [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis).
- The `X_idx_sorted` argument in [sksurv.tree.SurvivalTree.fit](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree.fit) has been deprecated in scikit-learn 0.24 and has no effect now.
- <span class="title-ref">predict\_cumulative\_hazard\_function</span> and <span class="title-ref">predict\_survival\_function</span> of [sksurv.ensemble.RandomSurvivalForest](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.ensemble.RandomSurvivalForest.html) and [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.tree.SurvivalTree.html) now return an array of [sksurv.functions.StepFunction](https://scikit-survival.readthedocs.io/en/v0.15.0/api/generated/sksurv.functions.StepFunction.html) objects by default. Use `return_array=True` to get the old behavior.
- Support for Python 3.6 has been dropped.
- Increase minimum supported versions of dependencies. We now require:

> | Package | Minimum Version |
> |--------------|-----------------|
> | Pandas | 0.25.0 |
> | scikit-learn | 0.24.0 |

This release features a complete overhaul of the [documentation](https://scikit-survival.readthedocs.io/en/v0.14.0/index.html). It features a new visual design, and the inclusion of several interactive notebooks in the [User Guide](https://scikit-survival.readthedocs.io/en/v0.14.0/user_guide/index.html).

In addition, it includes important bug fixes. It fixes several bugs in [sksurv.linear_model.CoxnetSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.linear_model.CoxnetSurvivalAnalysis.html#sksurv.linear_model.CoxnetSurvivalAnalysis) where `predict`, `predict_survival_function`, and `predict_cumulative_hazard_function` returned wrong values if features of the training data were not centered. Moreover, the score function of [sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis.html#sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis) and [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis) will now correctly compute the concordance index if `loss='ipcwls'` or `loss='squared'`.


Bug fixes

- [sksurv.column.standardize()](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.column.standardize.html#sksurv.column.standardize) modified data in-place. Data is now always copied.
- [sksurv.column.standardize()](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.column.standardize.html#sksurv.column.standardize) works with integer numpy arrays now.
- [sksurv.column.standardize()](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.column.standardize.html#sksurv.column.standardize) used biased standard deviation for numpy arrays (`ddof=0`), but unbiased standard deviation for pandas objects (`ddof=1`). It always uses `ddof=1` now. Therefore, the output, if the input is a numpy array, will differ from that of previous versions.
- Fixed [sksurv.linear_model.CoxnetSurvivalAnalysis.predict_survival_function()](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.linear_model.CoxnetSurvivalAnalysis.html#sksurv.linear_model.CoxnetSurvivalAnalysis.predict_survival_function) and [sksurv.linear_model.CoxnetSurvivalAnalysis.predict_cumulative_hazard_function()](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.linear_model.CoxnetSurvivalAnalysis.html#sksurv.linear_model.CoxnetSurvivalAnalysis.predict_cumulative_hazard_function), which returned wrong values if features of training data were not already centered. This adds an offset_ attribute that accounts for non-centered data and is added to the predicted risk score. Therefore, the outputs of `predict`, `predict_survival_function`, and `predict_cumulative_hazard_function` will be different to previous versions for non-centered data (139).
- Rescale coefficients of [sksurv.linear_model.CoxnetSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.linear_model.CoxnetSurvivalAnalysis.html#sksurv.linear_model.CoxnetSurvivalAnalysis) if `normalize=True`.
- Fix score function of [sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis.html#sksurv.ensemble.ComponentwiseGradientBoostingSurvivalAnalysis) and [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.14.0/api/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis) if `loss='ipcwls'` or `loss='squared'` is used. Previously, it returned `1.0 - true_cindex`.

Enhancements

- Add `sksurv.show_versions()` that prints the version of all dependencies.
- Add support for pandas 1.1
- Include interactive notebooks in documentation on readthedocs.
- Add user guide on [penalized Cox models](https://scikit-survival.readthedocs.io/en/v0.14.0/user_guide/coxnet.html).
- Add user guide on [gradient boosted models](https://scikit-survival.readthedocs.io/en/v0.14.0/user_guide/boosting.html).

This release fixes warnings that were introduced with 0.13.0.

Bug fixes

- Explicitly pass ``return_array=True`` in [sksurv.tree.SurvivalTree.predict](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree.predict) to avoid FutureWarning.
- Fix error when fitting [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree) with non-float dtype for time (127).
- Fix RuntimeWarning: invalid value encountered in true_divide in [sksurv.nonparametric.kaplan_meier_estimator](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.nonparametric.kaplan_meier_estimator.html#sksurv.nonparametric.kaplan_meier_estimator).
- Fix PendingDeprecationWarning about use of matrix when fitting [sksurv.svm.FastSurvivalSVM](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.FastSurvivalSVM.html#sksurv.svm.FastSurvivalSVM) if optimizer is `PRSVM` or `simple`.

The highlights of this release include the addition of [sksurv.metrics.brier_score](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.metrics.brier_score.html#sksurv.metrics.brier_score) and [sksurv.metrics.integrated_brier_score](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.metrics.integrated_brier_score.html#sksurv.metrics.integrated_brier_score) and compatibility with scikit-learn 0.23.

`predict_survival_function` and `predict_cumulative_hazard_function` of [sksurv.ensemble.RandomSurvivalForest](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.ensemble.RandomSurvivalForest.html#sksurv.ensemble.RandomSurvivalForest) and [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree) can now return an array of [sksurv.functions.StepFunction](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.functions.StepFunction.html), similar to [sksurv.linear_model.CoxPHSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.linear_model.CoxPHSurvivalAnalysis.html#sksurv.linear_model.CoxPHSurvivalAnalysis) by specifying ``return_array=False``. This will be the default behavior starting with 0.14.0.

Note that this release fixes a bug in estimating inverse probability of censoring weights (IPCW), which will affect all estimators relying on IPCW.

Enhancements

- Make build system compatible with PEP-517/518.
- Added [sksurv.metrics.brier_score](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.metrics.brier_score.html#sksurv.metrics.brier_score) and [sksurv.metrics.integrated_brier_score](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.metrics.integrated_brier_score.html#sksurv.metrics.integrated_brier_score) (101).
- [sksurv.functions.StepFunction](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.functions.StepFunction.html) can now be evaluated at multiple points in a single call.
- Update documentation on usage of `predict_survival_function` and
`predict_cumulative_hazard_function` (118).
- The default value of `alpha_min_ratio` of [sksurv.linear_model.CoxnetSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.linear_model.CoxnetSurvivalAnalysis.html#sksurv.linear_model.CoxnetSurvivalAnalysis) will now depend on the `n_samples/n_features` ratio. If ``n_samples > n_features``, the default value is 0.0001 If ``n_samples <= n_features``, the default value is 0.01.
- Add support for scikit-learn 0.23 (119).

Deprecations

- `predict_survival_function` and `predict_cumulative_hazard_function` of [sksurv.ensemble.RandomSurvivalForest](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.ensemble.RandomSurvivalForest.html#sksurv.ensemble.RandomSurvivalForest) and [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree) will return an array of [sksurv.functions.StepFunction](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.functions.StepFunction.html) in the future (as [sksurv.linear_model.CoxPHSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.linear_model.CoxPHSurvivalAnalysis.html#sksurv.linear_model.CoxPHSurvivalAnalysis) does). For the old behavior, use `return_array=True`.

Bug fixes

- Fix deprecation of importing joblib via sklearn.
- Fix estimation of censoring distribution for tied times with events. When estimating the censoring distribution, by specifying ``reverse=True`` when calling [sksurv.nonparametric.kaplan_meier_estimator](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.nonparametric.kaplan_meier_estimator.html#sksurv.nonparametric.kaplan_meier_estimator), we now consider events to occur before censoring. For tied time points with an event, those with an event are not considered at risk anymore and subtracted from the denominator of the Kaplan-Meier estimator. The change affects all functions relying on inverse probability of censoring weights, namely:
- [sksurv.nonparametric.CensoringDistributionEstimator](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.nonparametric.CensoringDistributionEstimator.html#sksurv.nonparametric.CensoringDistributionEstimator)
- [sksurv.nonparametric.ipc_weights](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.nonparametric.ipc_weights.html#sksurv.nonparametric.ipc_weights)
- [sksurv.linear_model.IPCRidge](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.linear_model.IPCRidge.html#sksurv.linear_model.IPCRidge)
- [sksurv.metrics.cumulative_dynamic_auc](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.metrics.cumulative_dynamic_auc.html#sksurv.metrics.cumulative_dynamic_auc)
- [sksurv.metrics.concordance_index_ipcw](https://scikit-survival.readthedocs.io/en/v0.13.0/generated/sksurv.metrics.concordance_index_ipcw.html#sksurv.metrics.concordance_index_ipcw)
- Throw an exception when trying to estimate c-index from uncomparable data (117).
- Estimators in ``sksurv.svm`` will now throw an exception when trying to fit a model to data with uncomparable pairs.

This release adds support for scikit-learn 0.22, thereby dropping support for older versions. Moreover, the regularization strength of the ridge penalty in [sksurv.linear_model.CoxPHSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.linear_model.CoxPHSurvivalAnalysis.html#sksurv.linear_model.CoxPHSurvivalAnalysis) can now be set per feature. If you want one or more features to enter the model unpenalized, set the corresponding penalty weights to zero. Finally, [sklearn.pipeline.Pipeline](https://scikit-learn.org/stable/modules/generated/sklearn.pipeline.Pipeline.html#sklearn.pipeline.Pipeline) will now be automatically patched to add support for `predict_cumulative_hazard_function` and `predict_survival_function` if the underlying estimator supports it.

Deprecations

- Add scikit-learn's deprecation of `presort` in [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree) and [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis).
- Add warning that default `alpha_min_ratio` in [sksurv.linear_model.CoxnetSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.linear_model.CoxnetSurvivalAnalysis.html#sksurv.linear_model.CoxnetSurvivalAnalysis) will depend on the ratio of the number of samples to the number of features in the future (41).

Enhancements

- Add references to API doc of [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis) (91).
- Add support for pandas 1.0 (100).
- Add `ccp_alpha` parameter for [Minimal Cost-Complexity Pruning](https://scikit-learn.org/stable/modules/tree.html#minimal-cost-complexity-pruning) to [sksurv.ensemble.GradientBoostingSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.ensemble.GradientBoostingSurvivalAnalysis.html#sksurv.ensemble.GradientBoostingSurvivalAnalysis).
- Patch [sklearn.pipeline.Pipeline](https://scikit-learn.org/stable/modules/generated/sklearn.pipeline.Pipeline.html#sklearn.pipeline.Pipeline) to add support for `predict_cumulative_hazard_function` and `predict_survival_function` if the underlying estimator supports it.
- Allow per-feature regularization for [sksurv.linear_model.CoxPHSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.linear_model.CoxPHSurvivalAnalysis.html#sksurv.linear_model.CoxPHSurvivalAnalysis) (102).
- Clarify API docs of [sksurv.metrics.concordance_index_censored](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.metrics.concordance_index_censored.html#sksurv.metrics.concordance_index_censored) (96).

This release adds [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree) and [sksurv.ensemble.RandomSurvivalForest](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.ensemble.RandomSurvivalForest.html#sksurv.ensemble.RandomSurvivalForest),
which are based on the log-rank split criterion. It also adds the OSQP solver as option to [sksurv.svm.MinlipSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.MinlipSurvivalAnalysis.html#sksurv.svm.MinlipSurvivalAnalysis) and [sksurv.svm.HingeLossSurvivalSVM](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.HingeLossSurvivalSVM.html#sksurv.svm.HingeLossSurvivalSVM), which will replace the now deprecated `cvxpy` and `cvxopt` options in a future release.

This release removes support for sklearn 0.20 and requires sklearn 0.21.

Deprecations

- The `cvxpy` and `cvxopt` options for `solver` in [sksurv.svm.MinlipSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.MinlipSurvivalAnalysis.html#sksurv.svm.MinlipSurvivalAnalysis) and
[sksurv.svm.HingeLossSurvivalSVM](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.HingeLossSurvivalSVM.html#sksurv.svm.HingeLossSurvivalSVM) are deprecated and will be removed in a future version. Choosing `osqp` is the preferred option now.

Enhancements

- Add support for pandas 0.25.
- Add OSQP solver option to [sksurv.svm.MinlipSurvivalAnalysis](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.MinlipSurvivalAnalysis.html#sksurv.svm.MinlipSurvivalAnalysis), and
[sksurv.svm.HingeLossSurvivalSVM](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.svm.HingeLossSurvivalSVM.html#sksurv.svm.HingeLossSurvivalSVM) which has no additional dependencies.
- Fix issue when using cvxpy 1.0.16 or later.
- Explicitly specify utf-8 encoding when reading README.rst (89).
- Add [sksurv.tree.SurvivalTree](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.tree.SurvivalTree.html#sksurv.tree.SurvivalTree) and [sksurv.ensemble.RandomSurvivalForest](https://scikit-survival.readthedocs.io/en/latest/generated/sksurv.ensemble.RandomSurvivalForest.html#sksurv.ensemble.RandomSurvivalForest) (90).

Bug fixes

- Exclude Cython-generated files from source distribution because
they are not forward compatible.