Removes duplicates from get_holidays output :param holidays_dict: dict(str, pd.DataFrame(EVENT_DF_DATE_COL, EVENT_DF_LABEL_COL)) dictiona

Basic descriptive stats on the timeseries time column. Returns ------- time_stats: `dict` Dictionary with descrip

Basic descriptive stats on the timeseries value column. Returns ------- value_stats : `dict` [`str`, `float`] Dic

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

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

Estimates the empirical distribution for a given variable in the ``distribution_col`` of ``df``. The distribution is an approximated quantile

Extracts forecasts, upper and lower confidence interval for each individual config. This is saved as a ``pandas.DataFrame`` with the name

Computes the column-wise anomaly scores, stored as ``score`` attribute. Parameters ---------- X : `pandas.DataFrame`

Fits the partial regularized regression. Parameters ---------- x : `numpy.array` or `pandas.DataFrame` The design

Runs ``forecast_pipeline`` on a rolling window basis. Parameters ---------- pipeline_params : `Dict` A dictionary containing the

Returns ``num`` of distinct RGB colors. If ``num`` is less than or equal to the length of ``colors``, first ``num`` elements of ``colors`` are

Returns the names of the ``.csv`` and ``.csv.xz`` files in ``data_path``. Parameters ---------- data_path : `str`

Get a changepoint dictionary based on the number of days in the observed timeseries and forecast horizon length in days to be provided as input to

Gets extra predictor columns from the model components for :func:`~greykite.framework.templates.silverkite_templates.silverkite_template`. Pa

Group-wise computation of forecasting error. Can be used to evaluate error/ aggregated value by a time feature, over time, or by a use

Gets the `ModelComponentsParam` class from model template. The template could be a name string, a `SimpleSilverkiteTemplateOptions` dataclass

Gets the number of potential changepoints for changepoint detection. Parameters ---------- n_points : `int` The number of data po

Returns rolling average runtime in seconds for ``config_names``. Parameters ---------- config_names : `list` [`str`], default

Given a list/tuple of strings (``strings``), it partitions it into various groups (sub-lists) as specified in a set of patterns given in ``regex_p

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

Loads the Hourly Parking dataset. This dataset contains occupancy rates (8:00 to 16:30) from 2016/10/04 to 2016/12/19 from car parks i

Computes the Matthews correlation coefficient for two arrays. The statistic is also known as the phi coefficient. The Matthews correlation coe

For a dataframe (``periods_df``) with rows denoting start and end of the periods, it plots the periods. If there extra segmentation is given (``gr

Plots group-wise evaluation metrics. Whereas `~greykite.framework.output.univariate_forecast.UnivariateForecast.plot_grouping_evaluation`

Returns a ``forecast_step`` ahead rolling forecast plot. The plot consists one line for each valid. ``config_names``. If available, th

Plots a Precision - Recall curve, where the x axis is recall and the y axis is precision. If ``grouping_col`` is None, it creates one Precision -

Predicts for new data. Parameters ---------- x : `numpy.array` or `pandas.DataFrame` The new data matrix.

This is the forecast function which can be used to forecast a number of periods into the future. It determines if the prediction shoul

Computes the prediction band width, expressed as a % relative to observed. Width is defined as average ratio of (upper-lower)/observed. :para

Returns quantile loss function for the specified quantile

Compute a precision score for two classes, usually labeled 1 and 0 to denote an anomaly and not an anomaly. Both ``y_true`` and ``y_pred`` need to

Compute a recall score for two classes, usually labeled 1 and 0 to denote an anomaly and not an anomaly. Both ``y_true`` and ``y_pred`` need to be

Simplifies a time feature predictor name. Eliminates the levels and unnecessary characters in a time feature predictor's name. If the origina

Computes the soft precision score for two classes, usually labeled 1 and 0 to denote an anomaly and not an anomaly. soft_precision(window) is

Computes the soft recall score for two classes, usually labeled 1 and 0 to denote an anomaly/ alert and not an anomaly/ alert, for `y_true` and `y

Splits pd.Dataframe(date, holiday) into separate dataframes, one per event Can be used to create the `daily_event_df_dict` parameter for `forecas

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

Imputes missing values in input time series. Checks the % of data points that are null, and provides warning if it exceeds ``self.max

This methods performs symmetric trimming from a given series `y` in the inputs. This means a percent of data from both sides of the

This function adds the prediction model features in training and predict phase for ``self.forecast`` internal use but can be called ou

Drops incomplete periods from the begin and end, and gets aggregated values. Calls ``self._drop_incomplete_agg`` with locations 0 and -1, the

Gets the non time columns in a df. Non time columns do not have f"{self.time_col}__" in it or do not equal to self.time_col. Paramete

If any parameters of the ``uncertainty_dict`` for specific uncertainty methods need to be populated from the estimator, they should be added h

Evaluates predicted target values for ``X`` relative to ``y_true``. Modified from `sklearn.metrics.scorer._PredictScorer` to work with

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

Returns a function that applies ``row_func`` to the selected ``cols``. Helper function for `~greykite.framework.output.univariate_forecast.Uni

Asserts for the default ForecastConfig values

Sweeps through ``map_func_dict``, converting values that are `~greykite.common.evaluation.ElementwiseEvaluationMetricEnum` member name

Sweeps through ``metric_dict``, converting members of ``enum_class`` to their corresponding evaluation function. For example::

Calculates the prediction coverages: prediction band width, prediction band coverage etc. :param observed: pd.Series or np.array, numeric, ob

Finds constant columns in x. A column is considered as a constant column if all rows have the same value and the value is not zero. Para

Creates a dictionary of predictor categories. The keys are categories, and the values are the corresponding predictor names. For detail, see

This is the main function to create dataframe with level shift regressor columns. It will detect the level shifts, return a dataframe

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

Identifies any gaps in pandas Series containing timestamps Timestamps are assumed to be in sorted order :return: pd.DataFrame with dates of th

Updates `self.impute_params`. Parameters ---------- X : `pandas.DataFrame` Training input data. e.g. each column

Fits the uncertainty model. Parameters ---------- train_df : `pandas.DataFrame` The data used to fit the uncertai

Formats the summary df for printing. Rounds numbers so they look good. Formats p-values to make them display the same way as the lm function

Computes the fraction of observed values between lower and upper. :param observed: pd.Series or np.array, numeric, observed values :param low

generates a data frame which includes a response variable with column name "y", and an estimate of y with column name "y_hat". y is designed to be

Generates daily event dict for all holidays inferred. The daily event dict will contain: - Single events for every holiday or hol

Automatically infers the following seasonality Fourier series orders: - yearly seasonality - quarterly seasonality - monthly

Group-wise computation of aggregated timeSeries value. Can be used to evaluate error/ aggregated value by a time feature, over time, o

Gets the explicit or implicit intercept column name from `patsy` design matrix. By default, `patsy` will make the design matrix always full rank.

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

Gets the seasonality change point feature df column names. Parameters ---------- df : `pandas.DataFrame` The dataframe used to bu

Gets the trend changepoint dates from trend changepoint column names. Parameters ---------- trend_cols : `list[`str`]` List of tr

Parses the adaptive lasso estimator to get change point dates. The functions calls ``adaptive_lasso_cv`` to selected potential trend change point

Groups ``df`` and evaluates a function on each group. The function takes a `pandas.DataFrame` and returns a scalar. Parameters ----------

Joins data with forecasts. This will be used both in training and prediction phases. Parameters ---------- forecast_df

Loads the Hourly Electricity dataset. This dataset contains the hourly consumption (in Kilowatt) of 321 clients from 2012 to 2014 pub

Loads the Hourly Parking dataset. This dataset contains occupancy rates (8:00 to 16:30) from 2016/10/04 to 2016/12/19 from car parks

Loads the Hourly San Francisco Bay Area Traffic dataset. This dataset contains the road occupancy rates (between 0 and 1) measured by

Loads the Hourly Solar Power dataset. This dataset contains the solar power production of an Australian wind farm from August 2019 to

Loads the Hourly Wind Power dataset. This dataset contains the wind power production of an Australian wind farm from August 2019 to J

Make a scorer from a performance metric or loss function. This factory function wraps scoring functions for use in GridSearchCV and cross_val

Calculates normal quantiles for each row of ``df`` using available means (either given in ``mean_col`` or ``fixed_mean``) and standard deviati

Parameters for ``plot_components``

Returns interactive plotly graph of the value against time. If anomaly info is provided, there is an option to show the anomaly adjustment.

Generic function to plot a dual y-axis plot. The x-axis is specified by ``x_col``. The left and right y-axes are specified by ``y_left_col`` and `

Returns a rolling plot of the forecasts by ``config_name`` against ``TIME_COL``. The plot consists of one line for each available split. Some

Computes error by group and plots the result. Can be used to plot error by a time feature, over time, or by a user-provided column. E

Computes aggregated timeseries by group and plots the result. Can be used to plot aggregated timeseries by a time feature, over time,

Returns a line plot of the grouped evaluation values of ``metric_dict`` of ``config_names``. These values are grouped by the grouping method c

Plots multiple lines against the same x-axis values. The lines can partially share the x-axis values. See parameter descriptions for a runnin

Simple plot of univariate timeseries. Parameters ---------- df : `pandas.DataFrame` Data frame with ``x_col`` and ``y_col`` x

Compare a collection by of timeseries (given in `df_dict`) by overlaying them in the specified period between ``start_time`` and ``end_t

Runs a pandas.DataFrame through hyperparameter_grid search with custom CV splits on a simple dataset to show that all the pieces fit together.

Simplifies an event predictor name. Eliminates the levels and unnecessary characters in an event predictor's name. If the original name is

Computes the soft F1 score for two classes, usually labeled 1 and 0 to denote an anomaly and not an anomaly. Soft F1 is simply calculated from

Calculates symmetric mean absolute percentage error Note that we do not include a factor of 2 in the denominator, so the range is 0% to 100%.

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