// Guides selection of the appropriate skforecast forecaster based on the user's data characteristics and requirements. Provides a decision matrix mapping use cases to forecaster classes. Use when the user is unsure which forecaster to use or asks for a recommendation.
Creates features for time series forecasting: calendar features with feature_engine (DatetimeFeatures, CyclicalFeatures), rolling statistics with RollingFeatures, differencing, and categorical exogenous variables. Use when the user wants to improve model accuracy through feature engineering or asks about exogenous variable creation.
Optimizes forecaster hyperparameters using grid search, random search, or Bayesian search (Optuna). Covers single-series and multi-series search, cross-validation configuration, and search space definition. Use when the user wants to find the best model configuration.
Diagnoses and fixes common errors when using skforecast, especially mistakes frequently made by LLMs generating skforecast code. Covers deprecated imports, wrong function names, missing parameters, and data format issues. Use when generated code produces errors or unexpected results.
Complete constructor signatures and method signatures for all skforecast forecasters, backtesting functions, search functions, cross-validation classes, preprocessing, feature selection, and drift detection. Use when the user needs exact parameter names, types, or defaults for any skforecast class or function.
Zero-shot time series forecasting with pre-trained foundation models (Amazon Chronos-2, Google TimesFM 2.5, Salesforce Moirai-2, Soda-INRIA TabICL) via ForecasterFoundation and FoundationModel. Covers single and multi-series workflows, exogenous variables, prediction intervals / quantiles, and backtesting. Use when the user wants forecasts without task-specific training, cold-start baselines, or pre-trained generalist models.
Generates prediction intervals for time series forecasts using bootstrapping, conformal prediction, or built-in statistical model intervals. Covers interval configuration, residual management, and calibration. Use when the user needs uncertainty quantification for forecasts.
| name | choosing-a-forecaster |
| description | Guides selection of the appropriate skforecast forecaster based on the user's data characteristics and requirements. Provides a decision matrix mapping use cases to forecaster classes. Use when the user is unsure which forecaster to use or asks for a recommendation. |
Use this skill when the user needs help choosing a forecaster, comparing forecaster types (recursive vs direct, single vs multi-series), or understanding which skforecast class fits their problem.
Skforecast is a machine learning-first library. The ML forecasters are the primary tools; statistical models (ForecasterStats) and naive baselines (ForecasterEquivalentDate) serve as comparison benchmarks.
| Scenario | Go to |
|---|---|
| 1 target series (with or without exogenous variables) | ā Step 2a |
| Multiple series to forecast simultaneously | ā Step 2b |
| Multiple series as drivers to predict one target | ā ForecasterDirectMultiVariate |
| Categorical target (e.g., low/medium/high) | ā ForecasterRecursiveClassifier |
| Scenario | Recommended Forecaster | Why |
|---|---|---|
| General purpose (start here) | ForecasterRecursive | Default choice. One model, recursive multi-step. Works with any sklearn-compatible estimator (LightGBM, XGBoost, CatBoost, RandomForest, etc.). Supports lags, window features, exog, differentiation, transformers, weight functions, and all probabilistic prediction methods (bootstrapping, conformal, quantiles, distributions) |
| Horizon-dependent patterns (e.g., predicting at 1h vs 24h requires different relationships) | ForecasterDirect | Trains one independent model per step ā no error propagation. Better when the predictive relationship changes significantly across the forecast horizon. Requires steps at init; parallelizable with n_jobs |
| Statistical baseline | ForecasterStats | Wraps ARIMA, SARIMAX, ETS, ARAR. Use as a benchmark to compare against ML models, or when the series is very short (< 200 obs) and ML overfits |
| Zero-shot / cold-start / no training data | ForecasterFoundation | Wraps pre-trained foundation models (Chronos-2, TimesFM 2.5, Moirai-2, TabICL). fit() only stores context ā no training. Good baseline and cold-start option. See the foundation-forecasting skill |
| Naive baseline | ForecasterEquivalentDate | Predicts using equivalent past dates (e.g., same weekday last week). Use as a sanity-check baseline |
| Scenario | Recommended Forecaster | Why |
|---|---|---|
| Forecast many series with a shared model (start here) | ForecasterRecursiveMultiSeries | One global model learns cross-series patterns. Supports DataFrame or dict input (dict allows series with different date ranges). Encoding options: 'ordinal' (default), 'ordinal_category', 'onehot', None. Supports per-series transformers, per-series differentiation, series_weights |
| Other series are features for one target | ForecasterDirectMultiVariate | All series become input features to predict a single level. Per-series lags via dict ({'sales': [1,7], 'price': [1]}). One model per step ā no error propagation |
| Deep learning / complex nonlinear patterns | ForecasterRnn | Keras-based RNN/LSTM/GRU. Single model outputs all steps and levels simultaneously via 3D tensors. Only conformal intervals (no bootstrapping). Requires keras |
| Zero-shot / pre-trained generalist | ForecasterFoundation | Global zero-shot forecasts via Chronos-2 / TimesFM 2.5 / Moirai-2 / TabICL. fit() only stores context. Native quantile intervals. Chronos-2 and TabICL support exog; TimesFM 2.5 & Moirai-2 do not. See the foundation-forecasting skill |
How many series?
ā
āāāŗ 1 series
ā ā
ā āāāŗ Is it a classification problem? āāāŗ Yes āāāŗ ForecasterRecursiveClassifier
ā ā
ā āāāŗ Regression (continuous target)
ā ā
ā āāāŗ Does the forecast relationship change across the horizon?
ā ā āāāŗ No / Unsure āāāŗ ForecasterRecursive ā START HERE
ā ā āāāŗ Yes (step-specific patterns) āāāŗ ForecasterDirect
ā ā
ā āāāŗ Compare with baselines:
ā āāāŗ ForecasterStats (Auto-ARIMA, ETS) as statistical benchmark
ā āāāŗ ForecasterEquivalentDate as naive benchmark
ā
āāāŗ Multiple series
ā
āāāŗ Want to forecast ALL series (global model)?
ā āāāŗ ForecasterRecursiveMultiSeries ā START HERE
ā
āāāŗ Want to use other series AS FEATURES for one target?
ā āāāŗ ForecasterDirectMultiVariate
ā
āāāŗ Need deep learning for very complex patterns?
āāāŗ ForecasterRnn
Zero-shot / no training data / cold-start (single or multi-series)?
āāāŗ ForecasterFoundation (Chronos-2 / TimesFM 2.5 / Moirai-2 / TabICL)
| Aspect | ForecasterRecursive | ForecasterDirect |
|---|---|---|
| Models trained | 1 | N (one per step) |
| Error propagation | Yes (predictions feed into next step) | No (each step uses only observed data) |
| Forecast horizon | Flexible (any steps at predict time) | Fixed at init (steps required) |
| Training speed | Fast (one model) | Slower (N models, parallelizable via n_jobs) |
| Memory | Lower (1 estimator) | Higher (N estimators stored in estimators_ dict) |
| Best for | Most cases, especially short-to-medium horizons | Long horizons where patterns change per step |
| Prediction intervals | Bootstrapping + conformal | Bootstrapping + conformal |
| Aspect | ForecasterRecursiveMultiSeries | ForecasterDirectMultiVariate |
|---|---|---|
| Goal | Forecast all series with one model | Use all series as features for one target |
| Input | DataFrame or dict of Series | DataFrame (all series as columns) |
| Target | All series (selected via levels) | One series (specified via level) |
| Strategy | Recursive (1 model, predictions feed back) | Direct (1 model per step, no error propagation) |
| Series identification | Encoding: 'ordinal', 'ordinal_category', 'onehot', None | All series create separate lag columns |
| Series with different ranges | Yes (via dict input) | No (all must share same range) |
| Per-series lags | No (same lags for all) | Yes (dict: {'sales': [1,7], 'price': [1]}) |
| Series weights | Yes (series_weights param) | No |
| Per-series transformers | Yes (transformer_series dict) | Yes (transformer_series dict) |
| Per-series differentiation | Yes (differentiation dict) | Yes (via differentiator_ per series) |
| Aspect | ML Forecasters | ForecasterStats |
|---|---|---|
| Primary role | Main forecasting tools | Comparison baseline |
| Best for | Medium-to-large datasets, complex patterns, many exogenous features | Short series, interpretability, parametric intervals |
| Estimators | Any sklearn-compatible (LightGBM, XGBoost, RF, etc.) | Arima, Sarimax, Ets, Arar |
| Lags / window features | Full support (lags, RollingFeatures) | No (model handles its own structure) |
| Differentiation | Built-in (differentiation param) | Handled within model (e.g., ARIMA d parameter) |
| Exogenous variables | Full support | Only SARIMAX |
| Prediction intervals | Bootstrapping (binned residuals) + conformal | Built-in parametric |
| Tuning | grid_search_forecaster, random_search_forecaster, bayesian_search_forecaster | grid_search_stats, random_search_stats |
| Backtesting | backtesting_forecaster / backtesting_forecaster_multiseries | backtesting_stats |
| Feature selection | select_features / select_features_multiseries | Not applicable |
| Feature | Recursive | Direct | RecursiveMultiSeries | DirectMultiVariate | Rnn | Stats | EquivalentDate | Classifier |
|---|---|---|---|---|---|---|---|---|
| Lags | ā | ā | ā | ā (per-series dict) | ā | ā | ā | ā |
Window features (RollingFeatures) | ā | ā | ā | ā | ā | ā | ā | ā |
| Exogenous variables | ā | ā | ā | ā | ā | SARIMAX only | ā | ā |
| Differentiation | ā | ā | ā (per-series) | ā (per-series) | ā | Within model | ā | ā |
| Transformer y/series | ā | ā | ā (per-series) | ā (per-series) | ā | ā | ā | ā |
| Transformer exog | ā | ā | ā | ā | ā | ā | ā | ā |
| Weight function | ā | ā | ā (per-series) | ā | ā | ā | ā | ā |
| Bootstrapping intervals | ā | ā | ā | ā | ā | ā | ā | ā |
| Conformal intervals | ā | ā | ā | ā | ā | ā | ā | ā |
| Binned residuals | ā | ā | ā | ā | ā | ā | ā | ā |
| Quantile predictions | ā | ā | ā | ā | ā | ā | ā | ā |
| Distribution fitting | ā | ā | ā | ā | ā | ā | ā | ā |
| Class probabilities | ā | ā | ā | ā | ā | ā | ā | ā |
| Feature importances | ā | ā | ā | ā | ā | ā | ā | ā |
Legend: ā = supported, ā = not supported/not applicable.
Once you have chosen a forecaster, follow these steps to get started:
ForecasterRecursive; multiple series ā ForecasterRecursiveMultiSeriesLGBMRegressor) is the best starting point ā fast, handles categoricals, good defaultsRollingFeatures (rolling mean, std, min, max) and DateTimeFeatureTransformer or create_datetime_features as exogenous variablesdifferentiation parameter instead of manual differencingbacktesting_forecaster + TimeSeriesFold for realistic multi-step evaluationbayesian_search_forecaster (Optuna-based) ā can include lags in the search spacepredict_interval(method='bootstrapping', use_binned_residuals=True) for uncertainty quantificationForecasterStats (Auto-ARIMA: Arima(order=None)) and ForecasterEquivalentDate to verify the ML model adds value