Issues

[Core] Partial-series updates fail with grouped target transforms

Nixtla/mlforecast#809 · opened

on Oct 24, 2026

1

cross_validation(refit=False, level=...) calibrates intervals on in-sample windows

Nixtla/mlforecast#805 · opened

on Oct 21, 2026

1

Transfer intervals with ids=: scale_aligned and error_scaled filter source scores by target ids

Nixtla/mlforecast#795 · opened

on Oct 15, 2026

Weighted transfer conformal fails when calibration features contain NaN

Nixtla/mlforecast#793 · opened

on Oct 14, 2026

2

Pooled keys: encode static key columns once per predict and grow update buckets once per update

Nixtla/mlforecast#785 · opened

on Oct 12, 2026

Allow predict(ids=...) and recursive fitted values whenever the id column is in every pooled key

Nixtla/mlforecast#784 · opened

on Oct 12, 2026

[Utils] Daily price generation fails across daylight-saving transitions

Nixtla/mlforecast#779 · opened

on Oct 12, 2026

update() misassigns values with pandas categorical ids in a different category order

Nixtla/mlforecast#776 · opened

on Oct 8, 2026

2

Weighted transfer conformal uses weight_col as a density-ratio feature

Nixtla/mlforecast#763 · opened

on Oct 5, 2026

[Models] Monte Carlo sampling-based probabilistic recursive reduction forecaster

Nixtla/mlforecast#749 · opened

on Sep 28, 2026

polars date features diverge from the pandas convention: 5 raise AttributeError, weekday silently returns 1-7

Nixtla/mlforecast#738 · opened

on Sep 9, 2026

1

Target transforms corrupt conformal calibration with refit=False cross-validation

Nixtla/mlforecast#733 · opened

on Sep 7, 2026

[Core] Allow NaN target values at `fit`/`preprocess`, dropping them only before model fitting

Nixtla/mlforecast#704 · opened

on Jul 29, 2026

Accept a one-row-per-id DataFrame for `static_features`

Nixtla/mlforecast#702 · opened

on Jul 24, 2026

3

Allow `predict(ids=...)` with safe pooled/grouped lag-transform subsets

Nixtla/mlforecast#649 · opened

on May 20, 2026

[Core] Support exogenous transforms inside `MLForecast` and allow raw exogenous inputs to be dropped from the model feature set

Nixtla/mlforecast#635 · opened

on Apr 28, 2026

[Core] Add interpretability capabilities trough SHAP

Nixtla/mlforecast#150 · opened

on Jun 8, 2023

5

MLForecast pipeline : Feature extraction/creation + Feature selection + Hyperparameter tuning

Feature

Nixtla/mlforecast#139 · opened

on May 20, 2023

2

Navigation Menu

Repository navigation

Issues

[Core] Partial-series updates fail with grouped target transforms

cross_validation(refit=False, level=...) calibrates intervals on in-sample windows

Transfer intervals with ids=: scale_aligned and error_scaled filter source scores by target ids

Weighted transfer conformal fails when calibration features contain NaN

Pooled keys: encode static key columns once per predict and grow update buckets once per update

Allow predict(ids=...) and recursive fitted values whenever the id column is in every pooled key

[Utils] Daily price generation fails across daylight-saving transitions

update() misassigns values with pandas categorical ids in a different category order

Weighted transfer conformal uses weight_col as a density-ratio feature

[Models] Monte Carlo sampling-based probabilistic recursive reduction forecaster

polars date features diverge from the pandas convention: 5 raise AttributeError, weekday silently returns 1-7

Target transforms corrupt conformal calibration with refit=False cross-validation

[Core] Allow NaN target values at `fit`/`preprocess`, dropping them only before model fitting

Accept a one-row-per-id DataFrame for `static_features`

Allow `predict(ids=...)` with safe pooled/grouped lag-transform subsets

[Core] Support exogenous transforms inside `MLForecast` and allow raw exogenous inputs to be dropped from the model feature set

[Core] Add interpretability capabilities trough SHAP

MLForecast pipeline : Feature extraction/creation + Feature selection + Hyperparameter tuning