Source code for interpreTS.core.features.feature_variance_change

import pandas as pd
import numpy as np



def calculate_change_in_variance(data, window_size=5):
    """
    Calculate the change in variance over time in a time series.

    Parameters
    ----------
    data : pd.Series or np.ndarray
        The time series data for which the change in variance is to be calculated.
    window_size : int, optional
        The size of the rolling window to calculate variance (default is 5).

    Returns
    -------
    pd.Series
        A series containing the change in variance over time, with the same index as the input.

    Raises
    ------
    TypeError
        If the data is not a valid time series type.
    ValueError
        If the data contains NaN values or is too short to calculate variance.

    Examples
    --------
    >>> import pandas as pd
    >>> data = pd.Series([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])
    >>> calculate_change_in_variance(data, window_size=3)
    0     NaN
    1     NaN
    2     NaN
    3     0.00
    4     0.00
    dtype: float64
    """
    # Convert data to a pandas Series if it's an ndarray
    if isinstance(data, np.ndarray):
        data = pd.Series(data)

    # Calculate rolling variance
    rolling_variance = data.rolling(window=window_size).var()

    # Calculate the change in variance as the first difference of rolling variances
    change_in_variance = rolling_variance.diff()

    return change_in_variance