Function reference

Data

Utilities for fetching and loading time series, weather, and holiday data.

data.fetch_data.fetch_data	Fetches a dataset from a CSV file or processes a DataFrame.
data.fetch_data.fetch_holiday_data	Fetches holiday data for the dataset period.
data.fetch_data.fetch_weather_data	Fetch weather data for the dataset period plus forecast horizon.
data.fetch_data.get_cache_home	Return the location where persistent models are to be cached.
data.fetch_data.get_data_home	Return the location where datasets are to be stored.
data.fetch_data.get_package_data_home	Return the location of the internal package datasets.
data.fetch_data.load_timeseries	Load the actual-load time series from interim/energy_load.csv.
data.fetch_data.load_timeseries_forecast	Load the day-ahead forecast time series from interim/energy_load.csv.
data.fetch_data.load_renewable_forecast	Load the ENTSO-E day-ahead wind/solar generation forecast.
data.fetch_data.load_day_ahead_price	Load the ENTSO-E day-ahead spot price (DE/LU) as an hourly series.
data.data_classes	Data structures for input and processed data.
data.demo_loader	Demo data loader for safety-critical forecasting tasks.
data.entsoe_loader	ENTSO-E interim-CSV data loaders.

Preprocessing

Hardened tools for data curation, resampling, outlier detection, feature engineering, and temporal train/test splitting.

preprocessing.coverage.assert_frontier_fresh	Raise CoverageError if the data frontier is stale.
preprocessing.coverage.assert_actual_lag_within	Raise CoverageError if the last published actual is too old.
preprocessing.coverage.assert_no_interior_gaps	Raise CoverageError if the recent actuals contain large holes.
preprocessing.coverage.last_complete_hour	Return the latest hour having a complete set of intra-hour samples.
preprocessing.curate_data.agg_and_resample_data	Aggregates and resamples the data to (e.g., hourly) frequency by computing the specified aggregation (e.g. for each hour).
preprocessing.curate_data.basic_ts_checks	Checks if the time series data has a datetime index and is sorted.
preprocessing.curate_data.curate_holidays	Checks if the holiday dataframe has the correct shape.
preprocessing.curate_data.curate_weather	Checks if the weather dataframe has the correct shape.
preprocessing.curate_data.get_start_end	Get start and end date strings for data and covariate ranges.
preprocessing.curate_data.remove_duplicate_timestamps	Resolve duplicate timestamps across all data columns.
preprocessing.curate_data.reset_index	Resets the index of the dataframe and assigns a name to the index column.
preprocessing.outlier.get_outliers	Detect outliers in each column using Isolation Forest.
preprocessing.outlier.manual_outlier_removal	Manual outlier removal function.
preprocessing.outlier.mark_outliers	Marks outliers as NaN in the dataset using Isolation Forest.
preprocessing.checking.check_exog	Validate that exog is a pandas Series or DataFrame.
preprocessing.checking.check_exog_dtypes	Check that exogenous variables have valid data types (int, float, category).
preprocessing.checking.check_interval	Validate that a confidence interval specification is valid.
preprocessing.checking.check_predict_input	Check all inputs of predict method. This is a helper function to validate
preprocessing.checking.check_residuals_input	Check residuals input arguments in Forecasters.
preprocessing.checking.check_y	Validate that y is a pandas Series without missing values.
preprocessing.checking.get_exog_dtypes	Extract and store the data types of exogenous variables.
preprocessing.checking.set_cpu_gpu_device	Set the device for the estimator to either ‘cpu’, ‘gpu’, ‘cuda’, or None.
preprocessing.exog_builder.ExogBuilder	Builds a set of exogenous features for a given date range.
preprocessing.exog_providers.ExogFeatureProvider	Contract for a pluggable exogenous-feature source.
preprocessing.exog_providers.CovidInfectionRateProvider	German national COVID-19 7-day incidence as an exogenous level regressor.
preprocessing.exog_providers.EntsoeForecastLoadProvider	ENTSO-E day-ahead Forecasted Load as an exogenous near-oracle prior.
preprocessing.exog_providers.EntsoeRenewableForecastProvider	ENTSO-E day-ahead wind and solar generation forecast.
preprocessing.exog_providers.EntsoeNetLoadProvider	ENTSO-E day-ahead net load = Forecasted Load − (wind + solar) forecast.
preprocessing.exog_providers.EntsoeDayAheadPriceProvider	ENTSO-E day-ahead spot price (DE/LU) as an exogenous input.
preprocessing.exog_providers.EventWindowProvider	Generic event-window provider driven by a bundled CSV file.
preprocessing.exog_providers.FootballMatchWindowProvider	German football match event-window provider.
preprocessing.exog_providers.EnergyCrisisWindowProvider	German energy-saving regulatory window provider.
preprocessing.exog_providers.build_providers	Construct the providers whose flags are truthy, in registry order.
preprocessing.exog_providers.build_providers_from_config	Construct providers by reading the registry flags off a config object.
preprocessing.imputation.apply_imputation	Apply imputation to a DataFrame based on the method specified in config.
preprocessing.imputation.WeightFunction	Callable class for sample weights that can be pickled.
preprocessing.target_corruption.TargetCorruptionReport	Immutable summary of a target-corruption detection and policy run.
preprocessing.target_corruption.detect_target_corruption	Detect physically-impossible target-column corruption in the native frame.
preprocessing.target_corruption.apply_target_corruption_policy	Apply the configured corruption policy to the native-cadence frame.
preprocessing.imputation.custom_weights	Return 0 if index is in or near any gap.
preprocessing.imputation.get_missing_weights	Return imputed DataFrame and a series indicating missing weights.
preprocessing.data_transform	Data transformation utilities for time series forecasting.
preprocessing.forecaster_config	Forecaster configuration utilities.
preprocessing.linearly_interpolate_ts	Linear interpolation transformer for time series data.
preprocessing.repeating_basis_function	Repeating Basis Function transformer for cyclical features.
preprocessing.rolling.RollingFeatures	Compute rolling window statistics over time series data.

Processing

Utilities for aggregated and n-to-n predictions.

processing.agg_predict.agg_predict	Aggregates multiple prediction columns into a single combined prediction series.
processing.n2n_predict.n2n_predict	End-to-end baseline forecasting using equivalent date method.
processing.n2n_predict_with_covariates.n2n_predict_with_covariates	End-to-end recursive forecasting with exogenous covariates.
processing.shape_check.ShapeCheckReport	Immutable result of a forecast shape plausibility check.
processing.shape_check.check_forecast_shape	Measure correlation and daily-range ratio between a forecast and its reference.

Forecaster

Recursive forecasting classes, seasonal baselines, and metrics.

forecaster.base	ForecasterBase class.
forecaster.recursive._forecaster_recursive.ForecasterRecursive	Recursive autoregressive forecaster for scikit-learn compatible estimators.
forecaster.recursive._forecaster_equivalent_date.ForecasterEquivalentDate	This forecaster predicts future values based on the most recent equivalent
forecaster.recursive._forecaster_recursive_multiseries
forecaster.metrics.add_y_train_argument	Add y_train argument to a function if it is not already present.
forecaster.metrics.calculate_coverage	Calculate coverage of a given interval.
forecaster.metrics.create_mean_pinball_loss	Create pinball loss for a given quantile.
forecaster.metrics.crps_from_predictions	Compute the Continuous Ranked Probability Score (CRPS) from predictions.
forecaster.metrics.crps_from_quantiles	Calculate the Continuous Ranked Probability Score (CRPS) from quantiles.
forecaster.metrics.mean_absolute_scaled_error	Mean Absolute Scaled Error (MASE).
forecaster.metrics.root_mean_squared_scaled_error	Root Mean Squared Scaled Error (RMSSE).
forecaster.metrics.symmetric_mean_absolute_percentage_error	Compute the Symmetric Mean Absolute Percentage Error (SMAPE).
forecaster.wrappers.model	Recursive forecaster model wrappers for different estimators.
forecaster.wrappers.lgbm	Recursive forecaster wrapper using LightGBM.
forecaster.wrappers.xgb	Recursive forecaster model wrappers for different estimators.

Splitter

Time-series-aware cross-validation folds and train/val/test holdout helpers.

splitter.split_base	Base class for time series cross-validation splitting.
splitter.split_one_step	One step ahead cross-validation splitting.
splitter.split_ts_cv	Time series cross-validation splitting.
splitter.split.split_abs_train_val_test	Splits a time series DataFrame into training, validation, and test sets based on absolute timestamps.
splitter.split.split_rel_train_val_test	Splits a time series DataFrame into training, validation, and test sets by percentages.
splitter.utils_common	Common validation and initialization utilities for model selection.

Backtesting

Backtesting orchestration for time series forecasters.

backtesting.validation.backtesting_forecaster	Backtesting of forecaster model following the folds generated by the TimeSeriesFold
backtesting.validation.backtesting_forecaster_one_step	Backtesting of forecaster model using one-step-ahead predictions.
backtesting.validation

Configurator

Task-level configuration objects for the forecasting pipelines.

configurator.config_demo.ConfigDemo	Configuration for the safety-critical demo task.
configurator.config_entsoe.ConfigEntsoe	Configuration for the ENTSO-E forecasting pipeline.
configurator.config_multi.ConfigMulti	Configuration for the multi-input forecasting pipeline.

Manager

High-level training, prediction, and model persistence orchestration.

manager.demo_metrics.calculate_metrics	Calculate MAE and MSE for numeric evaluation.
manager.trainer.get_last_model	Get the latest trained model from the cache.
manager.trainer.get_path_model	Yield the path to a model file for a given iteration and model name.
manager.trainer.load_iteration	Load a saved model at a given iteration.
manager.trainer.should_retrain	Decide whether a forecaster should be retrained.
manager.predictor.build_prediction_package	Build a prediction package for downstream plotting/reporting consumers.
manager.predictor.get_model_prediction	Get the prediction package from the latest trained model.
manager.logger	Audit-grade logging for spotforecast2-safe.
manager.persistence
manager.features.apply_cyclical_encoding	Apply cyclical (sine/cosine) encoding to periodic integer features.
manager.features.create_interaction_features	Append bilinear interaction terms to an exogenous feature matrix.
manager.features.select_exogenous_features	Select and deduplicate exogenous feature columns for model training.
manager.features.merge_data_and_covariates	Merge target data with exogenous features and split into train/predict slices.
manager.features.get_target_data	Extract the training series and exogenous slices for one target column.
manager.features.select_top_poly_features	Rank polynomial interaction columns by mutual information, keep the top K.

Multitask

Config-driven multi-target forecasting orchestrator. Tuning-free and plotting-free safe subset of the sibling spotforecast2.multitask. Provides BaseTask (shared pipeline steps), MultiTask (task dispatcher), task-specific classes (LazyTask, DefaultsTask, PredictTask, CleanTask), the module-level agg_predictor helper, a LightGBM factory, training strategies, and a one-call runner.

multitask.base.BaseTask	Shared base for all multi-target forecasting pipeline tasks.
multitask.base.agg_predictor	Aggregate per-target prediction packages into a weighted forecast.
multitask.multi.MultiTask	Orchestrates a multi-target time-series forecasting pipeline.
multitask.lazy.LazyTask	Task 1 — Lazy Fitting with default LightGBM parameters.
multitask.defaults.DefaultsTask	Task 2 — Defaults fitting (no tuning, no cached params).
multitask.predict.PredictTask	Task 5 — Predict-only using previously saved models.
multitask.clean.CleanTask	Cache-cleaning task — removes all cached data from the pipeline cache.
multitask.factories.default_lgbm_forecaster_factory	Return a fresh, unfitted LightGBM ForecasterRecursive.
multitask.factories.quantile_lgbm_forecaster_factory	Return one quantile-regression LightGBM ForecasterRecursive per quantile.
multitask.factories.predict_quantile_band	Assemble per-quantile forecasts into one non-crossing band.
multitask.guards.assert_no_leakage	Raise LeakageError if any forbidden column reached the model.
multitask.strategies.TrainingStrategy	Strategy interface for preparing a forecaster before the final fit.
multitask.strategies.LazyStrategy	Approach 1 — Lazy fitting with optional cached tuning.
multitask.strategies.DefaultsStrategy	Approach 2 — Train with defaults, no tuning, no cached params.
multitask.runner.run	Run the MultiTask forecasting pipeline and return predictions.

Utils

General-purpose utility functions: CPE generation, validation, data transforms, and generic TTL-aware atomic snapshot store.

utils.cpe.get_cpe_identifier	Generates the CPE 2.3 identifier for the spotforecast2-safe project.
utils.convert_to_utc	Utility functions for timezone conversion.
utils.parse.parse_bool	Parse case-insensitive boolean strings for CLI arguments.
utils.snapshot_store.SnapshotStore	Generic TTL-aware atomic snapshot store.
utils.snapshot_store.parse_snapshot_timestamp	Parse the UTC timestamp encoded in a snapshot filename stem.

Stats

Statistical primitives for time-series diagnostics. Pure compute, no plotting dependencies — safety-critical-friendly.

stats.stationarity.augmented_dickey_fuller	Run an Augmented Dickey-Fuller test on a time series.
stats.spectral.compute_periodogram	Compute the periodogram of a time series.
stats.spectral.PeriodogramResult	Container for the output of compute_periodogram().

Security

Masking and PII-protection utilities for safe logging (CWE-532, CWE-312).

security.masking.mask_estimator

Return a non-sensitive string representation of an estimator.

Calendar

Calendar, holiday, and day/night feature engineering for covariate construction.

calendar.holiday.create_holiday_df	Create a DataFrame with datetime index and a binary holiday indicator column.
calendar.holiday.get_holiday_features	Build public-holiday indicators and align them to a regular time grid.
calendar.holiday.create_holiday_adjacency_df	Create a DataFrame with binary adjacency indicators for public holidays.
calendar.holiday.get_holiday_adjacency_features	Build holiday-adjacency indicators and align them to a regular time grid.
calendar.holiday.create_day_type_df	Create a day-type refinement of the public-holiday column.
calendar.holiday.get_day_type_features	Build day-type indicators and align them to a regular time grid.
calendar.holiday.create_school_holiday_df	Create a DataFrame with a binary school-holiday indicator for a German state.
calendar.holiday.get_school_holiday_features	Build per-Bundesland school-holiday indicators and align them to a forecast grid.
calendar.features.get_calendar_features	Create calendar-based features for a contiguous time range.
calendar.features.get_day_night_features	Create day/night features using astronomical sunrise and sunset times.
calendar.features.get_ephemeris_features	Create continuous solar-geometry features from the ephemeris.

Weather

Weather data integration using the Open-Meteo API, derived weather features (degree-hours, apparent temperature, dew point), and population-weighted multi-city spatial aggregation.

weather.client.WeatherClient	Client for fetching weather data from Open-Meteo API.
weather.client.WeatherService	High-level service for weather data generation.
weather.features.get_weather_features	Fetch weather data and compute rolling-window features.
weather.derived.heating_degree_hours	Heating degree-hours :math:\max(base - T, 0).
weather.derived.cooling_degree_hours	Cooling degree-hours :math:\max(T - base, 0).
weather.derived.dew_point	Dew-point temperature via the Magnus-Tetens approximation.
weather.derived.apparent_temperature	Steadman apparent (“feels-like”) temperature.
weather.derived.population_weighted_average	Combine per-location weather frames into one demand-weighted index.
weather.derived.add_derived_weather_features	Append the requested derived columns to a raw weather frame (fail-safe).
weather.locations.WeatherLocation	A single weather sampling location with a population weight.
weather.locations.default_german_locations	Return the default population-weighted German load-centre registry.
weather.locations.coordinates	Extract (latitude, longitude) pairs in order.
weather.locations.weights	Extract the raw (un-normalised) weights in order.

Downloader

Data downloaders for external data sources (e.g. ENTSO-E).

downloader.entsoe.download_new_data	Download new load and forecast data from ENTSO-E.
downloader.entsoe.download_renewable_forecast	Download the ENTSO-E day-ahead wind/solar generation forecast.
downloader.entsoe.download_day_ahead_price	Download the ENTSO-E day-ahead spot price (DE/LU).
downloader.entsoe.merge_build_manual	Merge all raw CSV files from the ‘raw’ directory into a single interim file.
downloader.entsoe.download_zone_loads	Download Actual Total Load separately for each German TSO control area.
downloader.entsoe.assemble_zone_loads	Join the per-zone interim load files into one aligned, validated frame.
downloader.entsoe.ZoneResult	Structured result record for one zone in a download_zone_loads collect run.
downloader.entsoe.build_zone_qc_frame	Build a bottom-up QC frame from per-zone interim CSVs.

Exceptions

Custom exception types for safety-critical failure signalling.

exceptions	Custom exceptions and warnings for spotforecast2.
exceptions.CoverageError	Exception raised when operational data-coverage requirements are violated.
exceptions.LeakageError	Exception raised when forbidden columns are detected in model inputs.

Tasks

Executable tasks for demonstration and production pipelines.

tasks.task_safe_demo	Task demo: compare baseline, covariate, and custom LightGBM forecasts against ground truth.
tasks.task_safe_zone_load_demo	Task demo: four-zone bottom-up total-load forecast vs. a direct aggregate.