Releases | Open Source Science

https://github.com/nixtla/hierarchicalforecast - v1.2.1

Features

Support TopDownSparse and MiddleOutSparse for temporal disaggregation by @christophertitchen in https://github.com/Nixtla/hierarchicalforecast/pull/362

Bug fixes

Remove y requirement from reconcile on temporal aggregation by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/356
Fix TopDown for temporal reconciliation by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/359

Documentation

Minor fixes by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/349

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v.1.2.0...v1.2.1

- Python
Published by elephaint 11 months ago

https://github.com/nixtla/hierarchicalforecast - v.1.2.0

New Features

[FEAT] Cross-temporal reconciliation by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/309

Bug Fixes

[FIX]: ensure levels are sorted prior to adjacency matrix construction by @christophertitchen in https://github.com/Nixtla/hierarchicalforecast/pull/344

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v1.1.0...v.1.2.0

- Python
Published by elephaint 11 months ago

https://github.com/nixtla/hierarchicalforecast - v1.1.0

New features

[FEAT] Add sparse non-negative OLS and WLS via QP for MinTraceSparse by @christophertitchen in https://github.com/Nixtla/hierarchicalforecast/pull/319
[FEAT] Implement adjacency matrix by @christophertitchen in https://github.com/Nixtla/hierarchicalforecast/pull/332
[FEAT] Extremely fast forecast proportions by @christophertitchen in https://github.com/Nixtla/hierarchicalforecast/pull/334

Bug fixes

[FIX] Handle zero division in top down methods by @mattbuot in https://github.com/Nixtla/hierarchicalforecast/pull/325
[FIX] Raise warning on NaN values when using average proportions and proportion averages methods by @janrth in https://github.com/Nixtla/hierarchicalforecast/pull/335
[FIX] TopDown method failing on combinations with other methods by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/330
[FIX] ERM-reg and ERM-reg-bu equations by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/331
[FIX] Produce reproducable samples for PERMBU by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/337

Documentation

[FIX] API reference links, removal of unnecessary headers @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/318

New Contributors

@mattbuot made their first contribution in https://github.com/Nixtla/hierarchicalforecast/pull/325
@janrth made their first contribution in https://github.com/Nixtla/hierarchicalforecast/pull/335

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v1.0.1...v1.1.0

- Python
Published by elephaint 12 months ago

https://github.com/nixtla/hierarchicalforecast - v1.0.1

Hotfix

[FIX] Use Numpy bool_ instead of Python bool in eagerly compiled functions by @elephaint in https://github.com/Nixtla/hierarchicalforecast/pull/315

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v1.0.0...v1.0.1

- Python
Published by elephaint about 1 year ago

https://github.com/nixtla/hierarchicalforecast - v1.0.0

New features

[FEAT] Polars support in https://github.com/Nixtla/hierarchicalforecast/pull/305
[FEAT] Evaluation to utils in https://github.com/Nixtla/hierarchicalforecast/pull/311

Breaking changes

As of v1.0.0, HierarchicalForecast no longer supports the uniqueid as index. Users may have to perform a `.resetindex()` when using a Pandas DataFrame that has the unique_id still as index. The old behavior has been deprecated throughout the entire Nixtlaverse, so it may be wise to update all Nixtla packages to ensure the same consistent behavior is observed everywhere.

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v0.4.3...v1.0.0

- Python
Published by elephaint about 1 year ago

https://github.com/nixtla/hierarchicalforecast - v0.4.3

New Features

[FEAT] Sparse middle-out reconciliation via MiddleOutSparse @christophertitchen (#281)
[FEAT] Add support for exogenous variables in utils.aggregate @KuriaMaingi (#297)
[FEAT] Efficient Schafer-Strimmer for MinT @elephaint (#280)
[FEAT] Improve residuals-based reconciliation stability and faster ma.cov @elephaint (#295)

Dependencies

As of v0.4.3, hierarchicalforecast no longer officially supports Python 3.8, which is EOL.

- Python
Published by release-drafter[bot] over 1 year ago

https://github.com/nixtla/hierarchicalforecast - v0.4.2

New Features

Add sparse top-down reconciliation via TopDownSparse @christopher-titchen (#277)
Decrease wall time of _get_PW_matrices for BottomUp and BottomUpSparse @christopher-titchen (#276)
Efficient MinTrace (ols/wlsvar/wlsstruct/mintcov/mintshrink) @elephaint (#264)

Documentation

Create CODEOFCONDUCT.md @tracykteal (#267)
Fix evaluate argument in readme @jmoralez (#257)
Update ml frameworks example @jmoralez (#254)
Add step to trigger mintlify workflow @rpmccarter (#259)

Dependencies

Remove numpy pin @DManowitz (#272)

Enhancement

Warn when num_threads is not used in MinTrace @jmoralez (#273)

- Python
Published by release-drafter[bot] over 1 year ago

https://github.com/nixtla/hierarchicalforecast - v0.4.1

Bug Fixes

keep only observed combinations in aggregate @jmoralez (#242)

Documentation

Fix parsing errors and add mint.json @hahnbeelee (#243)

Enhancement

support parallel MinTrace optimization @jmoralez (#249)

- Python
Published by release-drafter[bot] over 2 years ago

https://github.com/nixtla/hierarchicalforecast - v0.4.0

New Features

Sparse Reconciliation @mcsqr (#210)
[FEAT] Probabilistic Forecasting Util Functions @dluuo (#195)
[FEAT] NeuralForecast Compatibility and Example Notebook @dluuo (#188)

Bug Fixes

fix aggregate function @jmoralez (#232)
[FIX] Aggregate unbalanced datasets @FedericoGarza (#190)
Fix assignment to unbound variable @nickto (#187)

Documentation

[Doc] Updated FavoritaComplete evaluation @kdgutier (#220)
[Doc] Added baseline version detail for replicability @kdgutier (#218)
[Doc] Added HierE2E Favorita baseline @kdgutier (#217)
[Doc] aggregate showdoc + external reconciliation tutorials' improvements @kdgutier (#214)
[Doc] First iteration of HierE2E baseline execution + Documentation detail improvements @kdgutier (#212)
[Doc] Added baseline experiments and minor protection to Normality reconciler @kdgutier (#203)
[FEAT] HierarchicalForecast With GluonTS Example Notebook @dluuo (#200)
[Doc] Fix intro installation typo @kdgutier (#193)

Enhancement

Fixes for large datasets @mcsqr (#229)
Rename MSSE into RelMSE, add new implementation of MSSE @nickto (#185)
[FEAT] Core Numeric Type and Null Protections @dluuo (#181)

- Python
Published by release-drafter[bot] over 2 years ago

https://github.com/nixtla/hierarchicalforecast - v0.3.0

Computational Efficiency Improvements

New aggregate function that generates the hierarchical time series and the aggregation constraints matrix. Improve from $O((N{a}+N{b})^{2}log(N{a}+N{b}))$ to $O((N{a}+N{b})$.
Vectorization of the creation of probabilistic prediction levels, before done in for loops now performed in a single vectorized numpy call.

Evaluation Utilities

Added scaled continuous ranked probability scores (sCRPS).
Added mean scaled squared errors (MSSE).
Added energy score metric.
Added random sampling outputs to probabilistic reconcilers.
Added core.bootstrap_reconcile method to apply over different random seeds the reconcilers and generate standard deviations.

Refactorization of the HierarchicalForecast classes

Overall improvement of the core.reconciliation method.
Decoupled the probabilistic reconciler classes from the mean reconciler classes.
Decoupled fit protections from reconciliation.
Reconciler's inputs now mostly receive mostly numpy arrays.
Simplified and deprecated dependencies.

Documentation Improvements

Installation guide.
New introduction tutorial with minimal, intuitive example.
Tutorial on evaluation of reconciliation probabilistic reconciliation baselines.

New Collaborators and HierarchicalForecast Paper

We started a fruitful collaboration with Souhaib Ben Taieb and Shanika Wickramasuriya.
We submitted the HierarchicalForecast library paper to the Journal of Machine Learning Research.

What's Changed

[FEAT] Ignore jupyter notebooks as part of languages in https://github.com/Nixtla/hierarchicalforecast/pull/120
[FEAT] Factorizing reverse_sigmah from HierarchicalReconciliation in https://github.com/Nixtla/hierarchicalforecast/pull/121
[FEAT] Decoupling _reconcile, from _get_PW_matrices. in https://github.com/Nixtla/hierarchicalforecast/pull/123
[FEAT] PW initialization in https://github.com/Nixtla/hierarchicalforecast/pull/124
Prob Reconciler's tests location in https://github.com/Nixtla/hierarchicalforecast/pull/125
Core Refactorization + Reconcilers.fit in https://github.com/Nixtla/hierarchicalforecast/pull/128
CircleCI in https://github.com/Nixtla/hierarchicalforecast/pull/129
Shared HReconciler + predict method in https://github.com/Nixtla/hierarchicalforecast/pull/131
[FEAT] Reconciler's sample method in https://github.com/Nixtla/hierarchicalforecast/pull/133
[FEAT] CRPS, MSSE and Energy Score metrics in https://github.com/Nixtla/hierarchicalforecast/pull/134
time tracking utils in https://github.com/Nixtla/hierarchicalforecast/pull/135
[FEAT] Faster creation of ProbReconciler's ordered levels in https://github.com/Nixtla/hierarchicalforecast/pull/137
[FIX] Matplotlib and numba errors in https://github.com/Nixtla/hierarchicalforecast/pull/142
[FIX] Circle ci integration in https://github.com/Nixtla/hierarchicalforecast/pull/141
[BUG] PERMBU unique_id order and num_samples in https://github.com/Nixtla/hierarchicalforecast/pull/143
[Bug] Fixed S_df categorical index ordering in https://github.com/Nixtla/hierarchicalforecast/pull/145
[FEAT] seed/num_samples usage possibility + MSSE evaluation example in https://github.com/Nixtla/hierarchicalforecast/pull/147
[FEAT] Faster aggregate function + Gaussian Log Score in https://github.com/Nixtla/hierarchicalforecast/pull/150
[FIX] Documentation + Update bib reference in https://github.com/Nixtla/hierarchicalforecast/pull/156
light improvements to readme in https://github.com/Nixtla/hierarchicalforecast/pull/157
[FIX] Use micromamba instead of miniconda (CI) in https://github.com/Nixtla/hierarchicalforecast/pull/167
[BUG] Added level domain protection for normality and permbu methods in https://github.com/Nixtla/hierarchicalforecast/pull/166
Level domain protection in https://github.com/Nixtla/hierarchicalforecast/pull/169
Omit expensive linear algebra when not necessary in MinTrace in https://github.com/Nixtla/hierarchicalforecast/pull/171
[FIX] Add correct github link in https://github.com/Nixtla/hierarchicalforecast/pull/173
[DOCS] Improved index, intro, quick start, and geographical forecasts in https://github.com/Nixtla/hierarchicalforecast/pull/175

New Contributors

@melopeo made their first contribution in https://github.com/Nixtla/hierarchicalforecast/pull/157
@mcsqr made their first contribution in https://github.com/Nixtla/hierarchicalforecast/pull/171
@cchallu made their first contribution in https://github.com/Nixtla/hierarchicalforecast/pull/175

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v0.2.1...v0.3.0

- Python
Published by FedericoGarza almost 3 years ago

https://github.com/nixtla/hierarchicalforecast - v0.2.1

What's Changed

Introduction tutorial in https://github.com/Nixtla/hierarchicalforecast/pull/102
[FIX] Docs source links in https://github.com/Nixtla/hierarchicalforecast/pull/107
[FIX] General plot_hierarchical_predictions_gap in https://github.com/Nixtla/hierarchicalforecast/pull/106
Doc: Updated ReadMe in https://github.com/Nixtla/hierarchicalforecast/pull/111
FEAT: add installation guide in https://github.com/Nixtla/hierarchicalforecast/pull/114
FEAT: Documentation Outline in https://github.com/Nixtla/hierarchicalforecast/pull/112
[FIX] Add correct link to StatsForecast in https://github.com/Nixtla/hierarchicalforecast/pull/115
[FIX] Deprecate mycolorpy dependency in https://github.com/Nixtla/hierarchicalforecast/pull/116
[FEAT] Add conda badge to readme in https://github.com/Nixtla/hierarchicalforecast/pull/117

Full Changelog: https://github.com/Nixtla/hierarchicalforecast/compare/v0.2.0...v0.2.1

- Python
Published by FedericoGarza about 3 years ago