[ENH] Add probabilistic forecasting support (predict_interval/predict_quantiles) to MOIRAIForecaster · sktime/sktime#9215

(12 comments) (1 reaction) (1 assignee)Python (7,162 stars) (1,192 forks)batch import

enhancementgood first issuemodule:forecasting

説明

Is your feature request related to a problem? Please describe. The MOIRAIForecaster currently only supports point forecasts via predict(). This issues aims to enable probabilistic forecasting in moirai forecaster. Requested by a sktime user in issue https://github.com/sktime/sktime/issues/8987#issuecomment-3450662542

Describe the solution you'd like We need to set tags accordingly, and implement methods like _predict_intervals or predict_quantiles for different data formats supported by the forecaster.

コントリビューターガイド

技術スタック: python
領域: datamachine learning
Issue 種別: feature
難度: 4
推定時間: over 1 week
活動状況: active
明確さ: clear
前提条件: probabilistic forecasting conceptsMOIRAIForecaster internalssktime forecasting interface
初心者向け度: 30
調査方針: Examine the existing MOIRAIForecaster implementation (likely in sktime/forecasting/moirai/). Study how other forecasters (e.g., AutoARIMA) implement predict intervals and predict quantiles. Review the referenced issue #8987 for user requirements. The solution involves setting tags and implementing methods for different data formats (pd.Series, pd.DataFrame, numpy). Check the forecaster's base class for required method signatures.