polynomial_decomposer
======================================================================================
.. py:module:: evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer
.. autoapi-nested-parse::
Component that removes trends from time series by fitting a polynomial to the data.
Module Contents
---------------
Classes Summary
~~~~~~~~~~~~~~~
.. autoapisummary::
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer
Contents
~~~~~~~~~~~~~~~~~~~
.. py:class:: PolynomialDecomposer(time_index: str = None, degree: int = 1, period: int = -1, random_seed: int = 0, **kwargs)
Removes trends and seasonality from time series by fitting a polynomial and moving average to the data.
Scikit-learn's PolynomialForecaster is used to generate the additive trend portion of the target data. A polynomial
will be fit to the data during fit. That additive polynomial trend will be removed during fit so that statsmodel's
seasonal_decompose can determine the addititve seasonality of the data by using rolling averages over the series'
inferred periodicity.
For example, daily time series data will generate rolling averages over the first week of data, normalize
out the mean and return those 7 averages repeated over the entire length of the given series. Those seven
averages, repeated as many times as necessary to match the length of the given target data, will be used
as the seasonal signal of the data.
:param time_index: Specifies the name of the column in X that provides the datetime objects. Defaults to None.
:type time_index: str
:param degree: Degree for the polynomial. If 1, linear model is fit to the data.
If 2, quadratic model is fit, etc. Defaults to 1.
:type degree: int
:param period: The number of entries in the time series data that corresponds to one period of a
cyclic signal. For instance, if data is known to possess a weekly seasonal signal, and if the data
is daily data, period should be 7. For daily data with a yearly seasonal signal, period
should be 365. Defaults to -1, which uses the statsmodels libarary's freq_to_period function.
https://github.com/statsmodels/statsmodels/blob/main/statsmodels/tsa/tsatools.py
:type period: int
:param random_seed: Seed for the random number generator. Defaults to 0.
:type random_seed: int
**Attributes**
.. list-table::
:widths: 15 85
:header-rows: 0
* - **hyperparameter_ranges**
- { "degree": Integer(1, 3)}
* - **invalid_frequencies**
- []
* - **modifies_features**
- False
* - **modifies_target**
- True
* - **name**
- Polynomial Decomposer
* - **needs_fitting**
- True
* - **training_only**
- False
**Methods**
.. autoapisummary::
:nosignatures:
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.clone
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.default_parameters
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.describe
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.determine_periodicity
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.fit
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.fit_transform
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.get_trend_dataframe
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.inverse_transform
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.is_freq_valid
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.load
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.parameters
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.plot_decomposition
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.save
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.set_period
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.transform
evalml.pipelines.components.transformers.preprocessing.polynomial_decomposer.PolynomialDecomposer.update_parameters
.. py:method:: clone(self)
Constructs a new component with the same parameters and random state.
:returns: A new instance of this component with identical parameters and random state.
.. py:method:: default_parameters(cls)
Returns the default parameters for this component.
Our convention is that Component.default_parameters == Component().parameters.
:returns: Default parameters for this component.
:rtype: dict
.. py:method:: describe(self, print_name=False, return_dict=False)
Describe a component and its parameters.
:param print_name: whether to print name of component
:type print_name: bool, optional
:param return_dict: whether to return description as dictionary in the format {"name": name, "parameters": parameters}
:type return_dict: bool, optional
:returns: Returns dictionary if return_dict is True, else None.
:rtype: None or dict
.. py:method:: determine_periodicity(cls, X: pandas.DataFrame, y: pandas.Series, acf_threshold: float = 0.01, rel_max_order: int = 5)
:classmethod:
Function that uses autocorrelative methods to determine the likely most signficant period of the seasonal signal.
:param X: The feature data of the time series problem.
:type X: pandas.DataFrame
:param y: The target data of a time series problem.
:type y: pandas.Series
:param acf_threshold: The threshold for the autocorrelation function to determine the period. Any values below
the threshold are considered to be 0 and will not be considered for the period. Defaults to 0.01.
:type acf_threshold: float
:param rel_max_order: The order of the relative maximum to determine the period. Defaults to 5.
:type rel_max_order: int
:returns:
The integer number of entries in time series data over which the seasonal part of the target data
repeats. If the time series data is in days, then this is the number of days that it takes the target's
seasonal signal to repeat. Note: the target data can contain multiple seasonal signals. This function
will only return the stronger. E.g. if the target has both weekly and yearly seasonality, the function
may return either "7" or "365", depending on which seasonality is more strongly autocorrelated. If no
period is detected, returns None.
:rtype: int
.. py:method:: fit(self, X: pandas.DataFrame, y: pandas.Series = None) -> PolynomialDecomposer
Fits the PolynomialDecomposer and determine the seasonal signal.
Currently only fits the polynomial detrender. The seasonality is determined by removing
the trend from the signal and using statsmodels' seasonal_decompose(). Both the trend
and seasonality are currently assumed to be additive.
:param X: Conditionally used to build datetime index.
:type X: pd.DataFrame, optional
:param y: Target variable to detrend and deseasonalize.
:type y: pd.Series
:returns: self
:raises NotImplementedError: If the input data has a frequency of "month-begin". This isn't supported by statsmodels decompose
as the freqstr "MS" is misinterpreted as milliseconds.
:raises ValueError: If y is None.
:raises ValueError: If target data doesn't have DatetimeIndex AND no Datetime features in features data
.. py:method:: fit_transform(self, X: pandas.DataFrame, y: pandas.Series = None) -> tuple[pandas.DataFrame, pandas.Series]
Removes fitted trend and seasonality from target variable.
:param X: Ignored.
:type X: pd.DataFrame, optional
:param y: Target variable to detrend and deseasonalize.
:type y: pd.Series
:returns:
The first element are the input features returned without modification.
The second element is the target variable y with the fitted trend removed.
:rtype: tuple of pd.DataFrame, pd.Series
.. py:method:: get_trend_dataframe(self, X: pandas.DataFrame, y: pandas.Series) -> list[pandas.DataFrame]
Return a list of dataframes with 4 columns: signal, trend, seasonality, residual.
Scikit-learn's PolynomialForecaster is used to generate the trend portion of the target data. statsmodel's
seasonal_decompose is used to generate the seasonality of the data.
:param X: Input data with time series data in index.
:type X: pd.DataFrame
:param y: Target variable data provided as a Series for univariate problems or
a DataFrame for multivariate problems.
:type y: pd.Series or pd.DataFrame
:returns:
Each DataFrame contains the columns "signal", "trend", "seasonality" and "residual,"
with the latter 3 column values being the decomposed elements of the target data. The "signal" column
is simply the input target signal but reindexed with a datetime index to match the input features.
:rtype: list of pd.DataFrame
:raises TypeError: If X does not have time-series data in the index.
:raises ValueError: If time series index of X does not have an inferred frequency.
:raises ValueError: If the forecaster associated with the detrender has not been fit yet.
:raises TypeError: If y is not provided as a pandas Series or DataFrame.
.. py:method:: inverse_transform(self, y_t: pandas.Series) -> tuple[pandas.DataFrame, pandas.Series]
Adds back fitted trend and seasonality to target variable.
The polynomial trend is added back into the signal, calling the detrender's inverse_transform().
Then, the seasonality is projected forward to and added back into the signal.
:param y_t: Target variable.
:type y_t: pd.Series
:returns:
The first element are the input features returned without modification.
The second element is the target variable y with the trend and seasonality added back in.
:rtype: tuple of pd.DataFrame, pd.Series
:raises ValueError: If y is None.
.. py:method:: is_freq_valid(cls, freq: str)
:classmethod:
Determines if the given string represents a valid frequency for this decomposer.
:param freq: A frequency to validate. See the pandas docs at https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#offset-aliases for options.
:type freq: str
:returns: boolean representing whether the frequency is valid or not.
.. py:method:: load(file_path)
:staticmethod:
Loads component at file path.
:param file_path: Location to load file.
:type file_path: str
:returns: ComponentBase object
.. py:method:: parameters(self)
:property:
Returns the parameters which were used to initialize the component.
.. py:method:: plot_decomposition(self, X: pandas.DataFrame, y: Union[pandas.Series, pandas.DataFrame], show: bool = False) -> Union[tuple[matplotlib.pyplot.Figure, list], dict[str, tuple[matplotlib.pyplot.Figure]]]
Plots the decomposition of the target signal.
:param X: Input data with time series data in index.
:type X: pd.DataFrame
:param y: Target variable data provided as a Series for univariate problems or
a DataFrame for multivariate problems.
:type y: pd.Series or pd.DataFrame
:param show: Whether to display the plot or not. Defaults to False.
:type show: bool
:returns:
The figure and axes that have the decompositions
plotted on them
(Multi series) dict[str, (matplotlib.pyplot.Figure, list[matplotlib.pyplot.Axes])]: A dictionary that maps the series id
to the figure and axes that have the decompositions plotted on them
:rtype: (Single series) matplotlib.pyplot.Figure, list[matplotlib.pyplot.Axes]
.. py:method:: save(self, file_path, pickle_protocol=cloudpickle.DEFAULT_PROTOCOL)
Saves component at file path.
:param file_path: Location to save file.
:type file_path: str
:param pickle_protocol: The pickle data stream format.
:type pickle_protocol: int
.. py:method:: set_period(self, X: pandas.DataFrame, y: pandas.Series, acf_threshold: float = 0.01, rel_max_order: int = 5)
Function to set the component's seasonal period based on the target's seasonality.
:param X: The feature data of the time series problem.
:type X: pandas.DataFrame
:param y: The target data of a time series problem.
:type y: pandas.Series
:param acf_threshold: The threshold for the autocorrelation function to determine the period. Any values below
the threshold are considered to be 0 and will not be considered for the period. Defaults to 0.01.
:type acf_threshold: float
:param rel_max_order: The order of the relative maximum to determine the period. Defaults to 5.
:type rel_max_order: int
.. py:method:: transform(self, X: pandas.DataFrame, y: pandas.Series = None) -> tuple[pandas.DataFrame, pandas.Series]
Transforms the target data by removing the polynomial trend and rolling average seasonality.
Applies the fit polynomial detrender to the target data, removing the additive polynomial trend. Then,
utilizes the first period's worth of seasonal data determined in the .fit() function to extrapolate
the seasonal signal of the data to be transformed. This seasonal signal is also assumed to be additive
and is removed.
:param X: Conditionally used to build datetime index.
:type X: pd.DataFrame, optional
:param y: Target variable to detrend and deseasonalize.
:type y: pd.Series
:returns:
The input features are returned without modification. The target
variable y is detrended and deseasonalized.
:rtype: tuple of pd.DataFrame, pd.Series
:raises ValueError: If target data doesn't have DatetimeIndex AND no Datetime features in features data
.. py:method:: update_parameters(self, update_dict, reset_fit=True)
Updates the parameter dictionary of the component.
:param update_dict: A dict of parameters to update.
:type update_dict: dict
:param reset_fit: If True, will set `_is_fitted` to False.
:type reset_fit: bool, optional