Transforms and evaluates on test data. Must call ``fit`` before calling this method. forecasts_test : `pandas.DataFrame`

Validates the inputs of a fourier series components dataframe called by ``SilverkiteEstimator`` to validate the input ``fs_components_df``.

Drops aggregations with incomplete periods. For example, a daily aggregation of hourly data will have a biased aggregation if the data starts

For a data frame of events with a time_col and label_col it adds shifted events prior and after the given events For example i

Fills the default values to ``model_components`` if not provided. Parameters ---------- model_components : :class:`~greykite.

Explicitly calls the method in `~greykite.framework.templates.base_template.BaseTemplate` to make use of the decorator in this class.

Explicitly calls the method in `~greykite.framework.templates.base_template.BaseTemplate` to make use of the decorator in this class.

This builds an reward function given an expected anomaly percent range. This is the expected percent of anomalies in the data by user. This is

Calculates quantiles corresponding to probs by simulating time series with specified parameters :param sim_df_func: callable a function wh

Calculates difference between two dataframes macthed on a given index. One intended application is to compare breakdown dfs from two related ML mo

Whether ``true_val`` is strictly between ``lower_val`` and ``upper_val``. Parameters ---------- true_val : float True value.

Estimates the seasonality effect with detected change points. Parameters ---------- df : `pandas.DataFrame` The data df. time

This is the main function to detect level shifts based on the z score threshold. Attributes ---------- df : `pandas.DataFrame

Fits, transforms, and evaluates on training data. Parameters ---------- forecasts : `pandas.DataFrame` Forecasts

Fits a basic forecast model which is aggregate based. As an example for an hourly time series we can assign the value of the most recent three wee

Returns blocks of indices to insert anomalies into. :param timeseries_length: int length of the time series into which we want to inject

Generates dataset for anomaly detection unit tests. Parameters ---------- freq: `str` Frequency of the dataset. periods: `int

Automatically gets the parameters for growth. Parameters ---------- df : `pandas.DataFrame` The input time series. time_col :

Returns list of available countries for modeling holidays :param countries: List[str] only look for available countries in this set :

Returns a list of holidays that occur any of the countries between the years specified. :param countries: List[str] countries for whi

Suitable for use as the `refit` parameter to `~sklearn.model_selection.RandomizedSearchCV`, after wrapping with `functools.partial`. Call

Cardinalty factor used to penalize the overlap size & positional reward. Parameters ---------- overlap_count : array-like, 1-D Ac

Gets the seasonality magnitude change arrays for each seasonality component from the estimated regression coefficients. In total there are::

Returns forecast time parameters. Uses ``self.df``, ``self.config``, ``self.regressor_cols``. Available parameters: - s

Gets the Fourier feature column names. Parameters ---------- df : `pandas.DataFrame` The input data. time_col : `str`

Computes the H matrix given ``x_mat`` and ``alpha`` for linear and ridge regression. The formula is ``H = inv(X.T @ X + alpha * np.eye(p)) @ X.T``

Returns hyperparameter grid. Implements the method in `~greykite.framework.templates.base_template.BaseTemplate`. Uses ``self.time_p

Gets the maximum autoregression order. Returns ------- max_ar_order : `int` The maximum autoregression order.

Generates model results summary. Parameters ---------- result : `ForecastResult` See :class:`~greykite.framework.pipeline.pipelin

Automatically detects seasonality change points. The function first converts changepoints_dict if "method" == "auto", then extracts trend cha

Imputes every column of ``df`` using `~greykite.common.features.timeseries_impute.impute_with_lags`. Parameters ---------- df : `pand

Implements the quantile regression with penalty terms. Supports sample weight and feature weights in penalty. This is solved with linear prog

Loads the Beijing Particulate Matter (PM2.5) dataset. https://archive.ics.uci.edu/ml/datasets/Beijing+PM2.5+Data This hourly data set

Loads the Hourly Bike Sharing Count dataset. This dataset contains aggregated hourly count of the number of rented bikes. The data al

Loads the Daily Bitcoin Transactions dataset. This dataset contains the number of Bitcoin transactions from 2009 to 2021. The dataset

Loads dataset by name from the internal data library. Parameters ---------- data_name : `str` Dataset to load fro

Implements the quantile regression without penalty terms. Supports sample weight. Applies the iterative re-weighted least square (IRLS) algor

Parameters for ``forecast_silverkite``

Returns a barplot of the train and test values of ``metric_dict`` of ``config_names``. Value of a metric for all ``config_names`` are plotted

Returns a barplot of the runtimes of ``config_names``. Parameters ---------- config_names : `list` [`str`], default None

Convenience function to plot the original trend changepoint detection results. Parameters ---------- params : `dict` or `None

A function which plots a values given in ``value_col`` of a dataframe ``df`` against x-axis values given in ``x_col`` and adds annotations based

Make a longterm avg plot by taking the average in a window and moving that across the time. That window can be a week, month, year etc

Creates forecast for dates specified in X The forecast one-by-one is supposed to forecast for the specified forecast horizon with the

Creates forecast for dates specified in X To enable caching, every subclass must call this at the beginning of its ``.predict()``. Be

Creates the content for printing the summary. Parameters ---------- info_dict : `dict` The output summary dictionary from

For a given timezone, it constructs a function which determines if a timestamp (`dt`) is inside the daylight saving period or not for a list o

Removes NAs from the input series. Also, importantly, will raise error if there are not enough data left (at least 2). Parameters

Calls :func:`~greykite.framework.pipeline.pipeline.forecast_pipeline` according to the ``json_str`` configuration parameters. Paramet

This function construct normal data and injects anomalies into the data (returned in ``"df"`` item of the result). It also creates two

Simplifies a changepoint predictor name. Changes "changepoint" to "cp" to shorten the changepoint features' names. If ``cst.CHANGEPOINT_COL_P

Simplifies predictor names in a list. Parameters ---------- pred_cols : `list` [ `str` ] A list of predictor names to be simplifi

Prints input parameters and Prophet model parameters. Returns ------- log_message : str log message printed to lo

Creates the model summary for the given model Parameters ---------- max_colwidth : `int` The maximum length for p

Replaces outliers with NaN. Parameters ---------- X : `pandas.DataFrame` Data to transform. e.g. each column is a

Normalizes data using the specified scaling method. Parameters ---------- X : `pandas.DataFrame` Data to transfor

Returns the unique dictionaries in the input list, preserving the original order. Replaces ``unique_elements_in_list`` because `dict` is not h

Builds autoregressive df to be used in forecast models. Parameters ---------- df : `pandas.Dataframe` Dataframe t

Builds lagged regressor df to be used in forecast models. Parameters ---------- df : `pandas.Dataframe` Dataframe

Given a single simple silverkite template's name, preprocess the name order and fill in default values. The naming of a simple silverkite tem

Computes various evaluation metrics for the forecast. :param data: pd.DataFrame with columns according to self.__init__ (e.g. sub

Gets config with default value for `model_template` and `model_components_param` if not provided. - model_template : default value is ``s

Gets the model templates for each sub-model. These templates are for ``self.get_regressor_cols`` and ``self.get_lagged_regressor_info``

Returns fourier series order from a `pandas.DataFrame` fourier series specification. Return value is capped by ``max_order``. Paramet

Extracts template class (e.g. `SimpleSilverkiteTemplate`) from the config. Currently only supports single templates in `~greykite.fram

print method

Print method.

print method.

Aggregates the ``df`` with the given ``agg_freq`` and applies the ``agg_func``. All columns whose names do not start with f"{time_col}__" wil

Builds a default ``self.x_mat`` if it is not given. Currently the class does not support customizing the built feature matrix. The bu

Given a dataframe, training/fitting configuration, gets the training data and fit data. If training data include incomplete periods during ag

Gets the aggregation function. Returns the input if it's None or a callable. Finds the corresponding callable from `~greykite

Calculates the average score for each event date. Returns ------- result : `dict` [`str`, `float`] A dictionary w

Computes the transformed Lasso input quantities. Parameters ---------- x : `numpy.array` The design matrix.

Gets the number of data points in a aggregation period. Parameters ---------- data_freq : `str` The data frequenc

Gets the lag based predictions. Parameters ---------- fut_df : `pandas.DataFrame` The input data that contains ``

Calculates the scores for a list of events, each with multiple occurrences. Returns ------- result : `dict` [`str`, `list` [`

Gets the AR order so that the model can use ``past_df`` to generate AR terms instead of imputation. This function only applies to the Silverk

Does pre-processing which includes: - adjust trend (optional) - aggregation (optional) Parameters --

Checks and overrides the input parameters, and makes sure the values are valid. Parameters ---------- configs : `list` [`Seas

Adds indicators columns to coefficient summary df for categaries. Parameters ---------- coef_summary : `pandas.DataFrame` The coe

Adds features to `joined_dfs` passed. Note that if nothing is passed, this will update ``self.joined_dfs`` Parameters -------

Decorator for `greykite.algo.reconcile.convex.reconcile_forecasts.ReconcileAdditiveForecasts.fit`. Fetches parameters based on ``method`` and call

Explicitly calls the method in `~greykite.framework.templates.base_template.BaseTemplate` to make use of the decorator in this class.

Implements template interface method. Takes input data and optional configuration parameters to customize the model. Returns a set of

Benchmarks silverkite template and returns the output as a list :param data_name: str Name of the dataset we are performing benchmarking

This is a calculation step which uses both ``data`` and ``param``. This is typically expected to be implemented by user. By default, i

Fits two types of uncertainty model depending on the input of conditional_cols

Create dataframe with additional regressor columns of the level shift detected. Parameters ---------- df : `pandas.DataFrame`

Expand dataframe to future dates based on the forecast_horizon. Parameters ---------- df : `pandas.DataFrame` The

Runs cross validation for all algo, data pairs. It returns a dictionary which contains the error for each model / data pair. The inpu

The absolute error between a single true and predicted value. Parameters ---------- true_val : float True value. pred_val : f

The quantile loss between a single true and predicted value. Parameters ---------- true_val : float True value. pred_val : fl

The residual between a single true and predicted value. Parameters ---------- true_val : float True value. pred_val : float

The absolute error between a single true and predicted value. Parameters ---------- true_val : float True value. pred_val : f