Makes predictions of prediction intervals for ``df`` based on the predictions and ``self.uncertainty_model``. Parameters ----

Computes the recall scores for two arrays. Parameters ---------- y_true : array-like, 1-D The actual categories. y_pred : arr

Adjusts the time series by removing trend. There methods to remove trend from time series are listed in `~greykite.algo.common.season

Decouples a list of {country}_{holiday} names into a list of (country, holiday) tuple or the other way around, depending on the input type.

Asserts equality between two instances of `ADConfig`. Raises a ValueError in case of a parameter mismatch. Parameters ---------- ad_c

Checks if `refit` is initialized with metric = expected_metric, and greater_is_better = expected_greater_is_better. Parameters

This function generates a function ("build_lags_func" in the returned dict) which when called builds a lag data frame and an aggregated lag da

A function which builds a dataframe including time series lags (in the form of data frame columns) for a given value column (value_col) from a

This method uses the fitted parameters to decide if a point should be labeled as outlier and also provides a score. This meth

Calls `~greykite.common.utils.python_utils.dictionary_values_to_lists` on the provided dictionary or on each item in a list of dictionaries.

Whether the relative difference between ``pred_val`` and ``true_val`` is strictly greater than ``rtol``. Parameters ---------- true_v

Looks for gaps in df[time_col] and returns a pandas.DataFrame with the missing rows added in. Warning: if freq doesn't match intended

Fits the uncertainty model with a given ``df`` and ``uncertainty_dict``. Parameters ---------- df : `pandas.DataFrame`

Computes anomaly dataframe from a labeled ``df``. Parameters ---------- df : `pandas.DataFrame` A data frame which includes minim

Returns a basic pipeline for univariate forecasting. Allows for outlier detection, normalization, null imputation, degenerate column removal,

Validates and merges overlapping anomaly periods in anomaly dataframe. Also standardizes column names. For example, consider the following in

Returns features for growth terms with continuous time origins at the changepoint_values (locations) specified Generates a time series fe

Applies the changepoint method specified in `changepoints_dict` to return the changepoint values :param changepoints_dict: Optional[Dict[str, any

Returns default forecast horizon, backtest, and cross-validation parameters, given the input frequency, size, and user requested values. This

Utility function to generate a suffix given an input ``date``. Parameters ---------- date : `datetime.datetime` Input timestamp.

Partitions interval into n_changepoints + 1 segments, placing a changepoint at left endpoint of each segment. The left most segment do

Generates the names of internal predictor columns from the event dictionary passed to `~greykite.algo.forecast.silverkite.forecast_silverkite.

Returns parameters for `greykite.algo.reconcile.convex.reconcile_forecasts.ReconcileAdditiveForecasts.fit` corresponding to recognized hierarc

Gets the hyperparameter grid for the Multistage Forecast Model. Returns ------- hyperparameter_grid : `dict` [`str`, `list` [

Utility function to generate a suffix given an input ``date``. Parameters ---------- date : `datetime.datetime` Input timestamp.

Can be used to set the default value in a dataclass to a mutable value. Provides a factory function that returns a copy of the provided argum

Plots predicted against actual. Parameters ---------- kwargs : additional parameters Additional parameters to pas

Creates forecast for dates specified in ``X``. Parameters ---------- X : `pandas.DataFrame` Input timeseries with

Creates forecast for the dates specified in ``X``. Currently does not support the regressor case where there is gap between train and

Performs predictions and calculates uncertainty using one simulation of future and calculate the error separately (not relying on mult

Processes the intercept term. Merges intercept in beta and dummy column in x if they are not there. Appends "(Intercept)" to the front of ``p

Calculates root mean square error. :param y_true: observed values given in a list (or numpy array) :param y_pred: predicted values given in a

Takes the absolute value of x and raises it to power of y. Then it multiplies the result by sign of x. This guarantees this function is non-d

Splits a pandas offset string into number part and frequency string part. Parameters ---------- offset_str : `str` An offset stri

Transforms the provided forecasts using the fitted ``self.transform_matrix``. Parameters ---------- forecasts_test :

Forms the constraints for convex optimization. The following attributes should be set before calling this method: - constraint_ma

Gets the event df for a single holiday. An event df has the format: pd.DataFrame({ "date": ["2020-09-01", "2021-09-01"],

For a list of records (each being a `dict`) and a set of parameters each having a list of potential values, it expands each record in all possible

Checks whether input arrays have equal length. :param arrays: one or more lists/numpy arrays/pd.Series :return: true if lengths are all equal

Sets default values for ``model_components``. Parameters ---------- model_components : :class:`~greykite.framework.templates.autogen.fore

Applies the default ModelComponentsParam values to the given object. Converts None to a ModelComponentsParam object. Unpacks a list o

Applies the default model template to the given object. Unpacks a list of a single element to the element itself. Sets default value

Asserts equality between two instances of `ForecastConfig`. Raises an error in case of parameter mismatch. Parameters ---------- fore

Checks if `scoring` has the expected keys and score functions defined by the other parameters. Parameters ---------- scoring : `dict`

Given a regression based ML model (``ml_model``) and a design matrix (``x_mat``), and a string based grouping rule (``grouping_regex_patterns_dict

A function to generate trend features from a given time series df. The trend features include columns of the format max(0, x-c_i), where c_i is t

Adds custom trend changepoints to detected trend changepoints. Compares the distance between custom changepoints and detected changepoints, a

Computes the fitted values with selected regressors indicated by ``regex`` Parameters ---------- x : `pandas.DataFrame` The desig

Computes the minimal index distance between two consecutive detected change points. Given a df, its time column, the number of change points that

Contaminate a dataframe with anomalies. If original value is y, the anomalous value is (1 +/- delta)y, the + or - chosen randomly. :param df:

Calculates correlation. :param y_true: observed values given in a list (or numpy array) :param y_pred: predicted values given in a list (or n

Given a formula it extracts the response vector (y) and builds the design matrix (x_mat). Parameters ---------- df : `pandas.DataFra

Given a dictionary, returns a copy whose values are either lists, distributions with a ``rvs`` method, or None. The output is suitable for h

Finds neighbor change points given their indices For example x = [1, 2, 3, 7, 8, 10, 15, 20] The return with `min_index_distance`=2 i

Fits and transforms training data. Parameters ---------- forecasts : `pandas.DataFrame` Forecasts to fit the adju

Flattens an array by removing 1 level of nesting. Parameters ---------- array : `list` [`list`] List of lists. Returns -

Converts a json string to the corresponding instance of the `ADConfig` class. Raises ValueError if the input is not a json string.

Parses list of dictionaries, applying `f` to the dictionary values. All items must be dictionaries.

Parses list of dictionaries or None elements, applying `f` to the dictionary values. If an element in the list is None, it is returned directly.

Applies a forecast function (`train_forecast_func`) to many derived timeseries from `df` which are moving windows of `df`. For each derived se

Gets the minimum distance between detected changepoints. Parameters ---------- min_increment : `datetime.timedelta` The minimum i

Extracts changepoints from changepoint configuration and input data :param df: pd.DataFrame Training data. A data frame which includes th

Gets the changepoint detection resample frequency parameter. Parameters ---------- n_points : `int` The number of data points in

Gets proper formatted strings for changepoint dates. The default format is "_%Y_%m_%d_%H". When necessary, it appends "_%M" or "_%M_%S". Par

Returns rolling train and test evaluation metric values. Parameters ---------- metric_dict : `dict` [`str`, `callable`]

Group-wise computation of evaluation metrics. Whereas ``self.get_grouping_evaluation`` computes one metric, this allows computation of any num

Runs model predictions on ``df`` and creates a `~greykite.framework.output.univariate_forecast.UnivariateForecast` object. Parameters ---

Returns splitwise rolling evaluation metric values. These values are grouped by the grouping method chosen by ``groupby_time_feature``,

Returns RandomizedSearchCV object for hyperparameter tuning via cross validation `sklearn.model_selection.RandomizedSearchCV` runs a full grid se

Extracts CV results from ``grid_search`` for the specified score function. Returns the correct ranks on the test set and a tuple of the scores acr

Returns regressor column names from the model components. Implements the method in `~greykite.framework.templates.base_template.BaseTemplate`

Returns regressor column names. Implements the method in `~greykite.framework.templates.base_template.BaseTemplate`. Returns

Provides ``scoring`` and ``refit`` parameters for `~sklearn.model_selection.RandomizedSearchCV`. Together, ``scoring`` and ``refit`` specify

Gets the yearly seasonality changepoint dates from change frequency. This is an internal function used for varying yearly seasonality effects in

Infers the most appropriate Fourier series order. Can infer multiple seasonality components with multiple configs at the same time. Th

Computes the Informedness also known as the Youden's J statistic for two arrays. Youden's J statistic is defined as J = sensitivity + specificity

This function operates on a given data frame (``df``) which includes time (given in ``time_col``) and metrics. For each metric (given in ``val

Calculates mean absolute percentage error. :param y_true: observed values given in a list (or numpy array) :param y_pred: predicted values gi

Calculates median absolute percentage error. :param y_true: observed values given in a list (or numpy array) :param y_pred: predicted values

Function that solves the following constrained optimization problem. maximize ``df``[``objective_col``] subject to ``df``[``constrain

The optimizer which picks the best possible parameter from the ones specified in ``param_iterable``, using the reward specified in the

Plots the transform matrix visually, as a grid. By default, negative values are red and positive values are blue. Parameters

Returns predictions and prediction intervals on new data using the machine-learning (ml) model fitted via ``fit_ml_model`` and the uncertainty

Calculates the quantile (pinball) loss with quantile q of y_true and y_pred. :param y_true: one column of true values. :type y_true: list or

Simplifies the predictor names to make them shorter. Parameters ---------- name : `str` The predictor name. Should be a simple t

Assuming dayligh saving starts on ``dst_start`` and ends on ``dst_end``, it performs a series of assertion to ensure the function ``is_dst_fun

For the corresponding parameter in each config in ``configs``, replaces `None` with default values; overrides all values if ``override

Checks that necessary input are provided in ``self.uncertainty_dict`` and ``self.train_df``. This check only raises `~greykite.sklearn

Forms the objective for convex optimization. The objective is a sum of four types of errors:: obj = adj + bias + train + var

Returns the keys from a `serialized_dict`, the output of `dumps`. Parameters ---------- serialized_dict: `dict` T

Adds vertical rectangle shadings to existing figure. Each vertical rectangle information is given in rows of data frame ``periods_df``. The in

Checks fitted hyperparameter grid search result. Parameters ---------- grid_search : `sklearn.model_selection.RandomizedSearchCV`

A function which returns a function to build autoregression dataframe for multiple value columns. This function should not be applied to data

A function which uses: df (pd.Dataframe), past_df (pd.DataFrame) and returns lag_df and agg_lag_df for df. This fu

Builds seasonality feature df from detected seasonality change points. The function is an extension of ``build_seasonality_feature_df_with_change