Parameters to tune the model.

Config for providing parameters to the Forecast library

The main entry point to create a forecast. Call the :py:meth:`~greykite.framework.templates.forecaster.Forecaster.run_forecast_config` method

How to split data for evaluation.

Returns datasets included in the library in `pandas.DataFrame` format. Attributes ---------- available_datasets : `list` [`str`]

Properties of the input data

Reward class which is to support very flexible set of rewards used in optimization where an objective is to be optimized. The main method for

A class to implement change point detection. Currently supports long-term change point detection only. Input is a dataframe with time_col ind

Wrapper for `~greykite.algo.forecast.silverkite.forecast_silverkite.SilverkiteForecast.forecast`. Parameters ---------- score_func : call

Flexible splitter for time-series cross validation and rolling window evaluation. Suitable for use in GridSearchCV. Attributes ----------

A derived class of `~greykite.algo.forecast.silverkite.SilverkiteForecast`. Provides an alternative interface with simplified configuration parame

Defines univariate time series input. The dataset can include regressors, but only one metric is designated as the target metric to forecast.

Config for providing parameters to the Anomaly Detection library.

Reconciles forecasts to satisfy additive constraints. Constraints can be encoded by the tree structure via ``levels``. In the tree formulatio

Wrapper for `~greykite.algo.forecast.silverkite.forecast_simple_silverkite.forecast_simple_silverkite`. Parameters ---------- score_func

A template for :class:`~greykite.sklearn.estimator.prophet_estimator.ProphetEstimator`. Takes input data and optional configuration parameters

A template for :class:`~greykite.sklearn.estimator.simple_silverkite_estimator.SimpleSilverkiteEstimator`. Takes input data and optional configur

Wrapper for Facebook Prophet model. Parameters ---------- score_func : callable see BaseForecastEstimator coverage : float b

The lag based estimator, using lagged observations with aggregation functions to forecast the future. This estimator includes the common week-over

A model template consists of a template class, a description, and a name. This class holds the template class and description. The model

The Multistage Forecast Estimator class. Implements the Multistage forecast method. The Multistage forecast method allows users to fit multip

Stores predicted and actual values. Provides functionality to evaluate a forecast: - plots true against actual with prediction bands.

This class is useful in constructing the data consumed in `~greykite.detection.detector.optimizer.Optimizer` class. Attributes --

This class is useful in constructing the data consumed in `~greykite.detection.detector.Detector` class. Attributes ----------

A base class for forecast estimators that fit using `~greykite.algo.forecast.silverkite.forecast_silverkite.SilverkiteForecast.forecast`. Not

What metrics to evaluate

Imputes nulls in time series data. This transform is stateless in the sense that ``transform`` output does not depend on the data passed to `

Simple selector that subsets a DataFrame to the columns of interest This allows a Pipeline that applies different transformations to subsets of co

How to compute the result.

Pandas Wrapper for DummyRegressor. Unlike DummyRegressor, it uses X for fitting instead of y Otherwise, the interface is identical. This allo

A dataclass to pass the parameters for `~greykite.algo.common.seasonality_inferrer.SeasonalityInferrer.infer_fourier_series_order`. Attribute

The dataclass to store Multistage forecast model config for a single model. Attributes ---------- train_length : `str`, default "392D"

Class for benchmarking multiple ForecastConfig on a rolling window basis. Attributes ---------- df : `pandas.DataFrame` Timeserie

Returns datasets included in the library in `pandas.DataFrame` or `~greykite.framework.input.univariate_time_series.UnivariateTimeSeries` format.

This is a dataclass to denote the result of an outlier detection.

This class enables Greykite based anomaly detection algorithms. It takes a ``forecast_config`` and ``ad_config`` (see Parameters) and builds a

An error that should be raised during fitting/prediction uncertainty. This type of error will be caught and won't fail the entire pipeline, be

Implements methods to automatically infer holiday effects. The class works for daily and sub-daily data. Sub-daily data is aggregated into da

Class that applies default values to a `~greykite.framework.templates.autogen.forecast_config.ForecastConfig` object. Provides these methods:

Extends the functionality of dill to serialize arbitrary objects. Originally intended to serialize the anomaly detector class `~greykite.dete

Simple estimator that predicts a constant (the weekday of the first date in ``X`` passed to ``predict``). Used to test BaseForecastEstimator m

The model template for Multistage Forecast Estimator.

Wrapper for ``pmdarima.arima.AutoARIMA``. It currently does not handle the regressor issue when there is gap between train and predict periods

Represents hierarchical relationships between nodes (time series). Nodes are indexed by their position in the tree, in breadth-first search (BFS)

Implements the quantile regression model. Supports weighted sample, l1 regularization and weighted l1 regularization. These options can b

The level shifts are handled with regressors in the Silverkite forecasting model. This class can generate corresponding regressors to the input da

Defines generic simple silverkite template options. Attributes can be set to different values using `~greykite.framework.templates.simple_sil

Simple selector that subsets a DataFrame to the columns of interest by their type This allows a Pipeline that applies different transformations to

Contains y-values for related lines to plot, such as forecasts or actuals. The lines share the same color, name, and legend group.

Forecast one-by-one estimator. Parameters ---------- score_func : `callable`, default mean_squared_error Function to calculate mo

A class to infer appropriate Fourier series orders in different seasonality components. The method allows users to: - optionally rem

A template for :class:`~greykite.sklearn.estimator.silverkite_estimator.SilverkiteEstimator`. Takes input data and optional configuration paramet

The simple conditional residuals uncertainty model. For more details, see `~greykite.algo.uncertainty.conditional.conf_interval`. Attributes

A template for :class:`~greykite.sklearn.estimator.auto_arima_estimator.AutoArimaEstimator`. Takes input data and optional configuration paramete

Replaces outliers in data with NaN. Outliers are determined by anomaly scores computed by the differences or ratios between the observed value

The quantile regression based uncertainty model. Predicts quantiles as the prediction intervals. The quantiles are calculated based on the ``c

This is a class for detecting one-dimensional outliers using z-score (based on the normal distribution). See https://en.wikipedia.org/wiki/Standar

Replaces outliers in data with NaN. Outliers are determined by z-score cutoff. Columns are handled independently. Parameters ----------

This data class represents the standard return of the method: `calc_with_param` of `~greykite.detection.detector.optimizer.Optimizer` Attribu

The dataclass to store Multistage Forecast model config for a single model. Attributes ---------- train_length : `str`, default "392D"

Provides properties for modeling for various time frequencies in Silverkite.

This is a class for detecting one-dimensional outliers. This uses the celebrated outlier definition of John Tukey (and named here as such in his r

Class to implement a partially regularized regression. A partially regularized regression is defined as beta_0, beta_1, beta_2 = argmin

This is the base class for detecting one-dimensional outliers. These classes are expected to return (outlier) scores for each point and a

This dataclass is to denote a `diff_method` if a differencing with respect to a baseline is needed.

Normalizes time series data. Parameters ---------- normalize_algorithm : `str` or None, default None Which algorithm to use. Vali

Contains information to create `fs_components_df` parameter for `forecast_silverkite` for modeling seasonality.

This module estimates the impact of holidays and their neighboring days given a raw holiday dataframe ``holiday_df``, and a time series containing

A template for :class: `~greykite.sklearn.estimator.lag_based_estimator.LagBasedEstimator`.

Concatenates results of multiple transformer objects. Transformers are expected to have pd.DataFrame as input and output Modified from sklear

Defines groupby time feature, xlabel and ylabel for Silverkite Component Plots.

Removes degenerate (constant) columns. Parameters ---------- drop_degenerate : `bool`, default False Whether to drop degenerate c

The base uncertainty model. Attributes ---------- uncertainty_dict : `dict` [`str`, any] The uncertainty model specification. It

A class to enable easy implementation of optimization over arbitrary parameter spaces and using arbitrary rewards. The optimization problem ca

Implements the cross-validation version of `~greykite.algo.common.partial_regularize_regression.PartialRegularizeRegression`. Attributes

Constants used by `~greykite.framework.templates.simple_silverkite_template.SimpleSilverkiteTemplate`. Includes the model templates and their

Bootstrap generator, an iterable object. The generator for bootstrap samples. Iterating along the object will give the bootstrap sample indic

Calculates time series features (e.g. year, month, hour etc.) of the input time series Parameters ---------- time_col : string, default=T

Base detector class for Anomaly Detection. The class initializes by passing an arbitrary ``reward`` for optimization and a potentially multiva

A class to store regression model summary statistics. The class can be printed to get a well formatted model summary. Attributes -------

Custom constants that will be used by Silverkite.

A detector based on normal distribution. A normal distribution if fitted to data and then any points outside the range ``(

The data class to store uncertainty models. Attributes ---------- model_class : `~greykite.sklearn.uncertainty.base_uncertainty_model.Bas

This class implements APE (absolute percent error) based detector. The class finds the - best forecast among multiple forecasts which can be

This class purpose is to find the best forecast from given k forecasts to act as baseline for anomaly detection. This class inherits its para

This class enables anomaly detection algorithms which use baseline forecasts in their logic. The class assumes that for a given dataset (`df`

This class is useful in constructing the data consumed in `~greykite.detection.detector.forecast_based.ForecastBasedDetector` Attributes

Forecast results. Contains results from cross-validation, backtest, and forecast, the trained model, and the original input data.

Constants used by `~greykite.framework.templates.multistage_forecast_template.MultistageForecastTemplate`. Include the model templates and the

Uses the appropriate constant mixins to provide all the constants that will be used by Silverkite.

Original class: `from plotly.io._base_renderers import SphinxGalleryRenderer` Modified to add `render_png` parameter.

A detector based on Tukey's outliar definition which is used in Boxplots (also invented by Tukey) as well to draw the whiskers. Refere