Rolling Origin Recalibration method

timecave.validation_methods.OOS.RollingOriginRecalibration(ts, fs=1, origin=0.7)

Bases: BaseSplitter

Implements the Rolling Origin Recalibration method.

This class implements the Rolling Origin Recalibration method. This method splits the data into various training and validation sets. Neither the training sets nor the validation sets are disjoint. At every iteration, a single data point is dropped from the validation set and added to the training set.

Parameters:

Name	Type	Description	Default
ts	ndarray | Series	Univariate time series.	required
fs	float | int	Sampling frequency (Hz).	1
origin	int | float	The point from which the data is split. If an integer is passed, it is interpreted as an index. If a float is passed instead, it is treated as the percentage of samples that should be used for training.	0.7

Attributes:

Name	Type	Description
n_splits	int	The number of splits.
sampling_freq	int | float	The series' sampling frequency (Hz).

Methods:

Name	Description
split	Split the time series into training and validation sets.
info	Provide additional information on the validation method.
statistics	Compute relevant statistics for both training and validation sets.
plot	Plot the partitioned time series.

Raises:

Type	Description
TypeError	If origin is neither an integer nor a float.
ValueError	If origin is a float that does not lie in the ]0, 1[ interval.
ValueError	If origin is an integer that does not lie in the ]0, n_samples[ interval.

Warning

Depending on the time series' size, this method can have a large computational cost.

See also

Rolling Origin Update: Similar to the Rolling Origin Recalibration method, but the model is only trained once.

Notes

The Rolling Origin Recalibration method consists of splitting the data into a training set and a validation set, with the former preceding the latter. At every iteration, a single data point (the one closest to the training set) is dropped from the validation set and added to the training set. The model's performance is then assessed on the new validation set. This process ends once the validation set consists of a single data point. The estimate of the true model error is the average validation error across [over] all iterations.

For more details on this method, the reader should refer to [1].

References

1

Leonard J Tashman. Out-of-sample tests of forecasting accuracy: an analysis and review. International journal of forecasting, 16(4):437–450, 2000.

Source code in timecave/validation_methods/OOS.py

def __init__(
    self, ts: np.ndarray | pd.Series, fs: float | int = 1, origin: int | float = 0.7
) -> None:

    super().__init__(2, ts, fs)
    self._check_origin(origin)
    self._origin = self._convert_origin(origin)
    self._splitting_ind = np.arange(self._origin + 1, self._n_samples)
    self._n_splits = self._splitting_ind.shape[0]

    return

info()

Provide some basic information on the training and validation sets.

This method displays the minimum and maximum sizes for both the training and validation sets.

Examples:

>>> import numpy as np
>>> from timecave.validation_methods.OOS import RollingOriginRecalibration
>>> ts = np.ones(10);
>>> splitter = RollingOriginRecalibration(ts);
>>> splitter.info();
Rolling Origin Recalibration method
-----------------------------------
Time series size: 10 samples
Minimum training set size: 7 samples (70.0 %)
Maximum validation set size: 3 samples (30.0 %)
Maximum training set size: 9 samples (90.0 %)
Minimum validation set size: 1 samples (10.0 %)

Source code in timecave/validation_methods/OOS.py

def info(self) -> None:
    """
    Provide some basic information on the training and validation sets.

    This method displays the minimum and maximum sizes for both the training and validation sets.

    Examples
    --------
    >>> import numpy as np
    >>> from timecave.validation_methods.OOS import RollingOriginRecalibration
    >>> ts = np.ones(10);
    >>> splitter = RollingOriginRecalibration(ts);
    >>> splitter.info();
    Rolling Origin Recalibration method
    -----------------------------------
    Time series size: 10 samples
    Minimum training set size: 7 samples (70.0 %)
    Maximum validation set size: 3 samples (30.0 %)
    Maximum training set size: 9 samples (90.0 %)
    Minimum validation set size: 1 samples (10.0 %)
    """

    max_training_size = self._n_samples - 1
    min_training_size = self._origin + 1
    max_validation_size = self._n_samples - self._origin - 1
    min_validation_size = 1

    max_training_pct = np.round(max_training_size / self._n_samples, 4) * 100
    min_training_pct = np.round(min_training_size / self._n_samples, 4) * 100
    max_validation_pct = np.round(max_validation_size / self._n_samples, 4) * 100
    min_validation_pct = np.round(min_validation_size / self._n_samples, 4) * 100

    print("Rolling Origin Recalibration method")
    print("-----------------------------------")
    print(f"Time series size: {self._n_samples} samples")
    print(
        f"Minimum training set size: {min_training_size} samples ({min_training_pct} %)"
    )
    print(
        f"Maximum validation set size: {max_validation_size} samples ({max_validation_pct} %)"
    )
    print(
        f"Maximum training set size: {max_training_size} samples ({max_training_pct} %)"
    )
    print(
        f"Minimum validation set size: {min_validation_size} samples ({min_validation_pct} %)"
    )

    return

plot(height, width)

Plot the partitioned time series.

This method allows the user to plot the partitioned time series. The training and validation sets are plotted using different colours.

Parameters:

Name	Type	Description	Default
height	int	The figure's height.	required
width	int	The figure's width.	required

Examples:

>>> import numpy as np
>>> from timecave.validation_methods.OOS import RollingOriginRecalibration
>>> ts = np.arange(1, 11);
>>> splitter = RollingOriginRecalibration(ts);
>>> splitter.plot(10, 10);

Source code in timecave/validation_methods/OOS.py

def plot(self, height: int, width: int) -> None:
    """
    Plot the partitioned time series.

    This method allows the user to plot the partitioned time series. The training and validation sets are plotted using different colours.

    Parameters
    ----------
    height : int
        The figure's height.

    width : int
        The figure's width.

    Examples
    --------
    >>> import numpy as np
    >>> from timecave.validation_methods.OOS import RollingOriginRecalibration
    >>> ts = np.arange(1, 11);
    >>> splitter = RollingOriginRecalibration(ts);
    >>> splitter.plot(10, 10);

    ![Holdout_plot_image](../../../images/RollRecal_plot.png)
    """

    fig, axs = plt.subplots(self._n_samples - self._origin - 1, 1, sharex=True)
    fig.set_figheight(height)
    fig.set_figwidth(width)
    fig.supxlabel("Samples")
    fig.supylabel("Time Series")
    fig.suptitle("Rolling Origin Recalibration method")

    for it, (training, validation, _) in enumerate(self.split()):

        axs[it].scatter(training, self._series[training], label="Training set")
        axs[it].scatter(
            validation, self._series[validation], label="Validation set"
        )
        axs[it].set_title("Iteration {}".format(it + 1))
        axs[it].legend()

    plt.show()

    return

split()

Split the time series into training and validation sets.

This method splits the series' indices into disjoint sets containing the training and validation indices. At every iteration, an array of training indices and another one containing the validation indices are generated. Note that this method is a generator. To access the indices, use the next() method or a for loop.

Yields:

Type	Description
ndarray	Array of training indices.
ndarray	Array of validation indices.
float	Used for compatibility reasons. Irrelevant for this method.

Examples:

>>> import numpy as np
>>> from timecave.validation_methods.OOS import RollingOriginRecalibration
>>> ts = np.ones(10);
>>> splitter = RollingOriginRecalibration(ts);
>>> for ind, (train, val, _) in enumerate(splitter.split()):
...
...     print(f"Iteration {ind+1}");
...     print(f"Training set indices: {train}");
...     print(f"Validation set indices: {val}");
Iteration 1
Training set indices: [0 1 2 3 4 5 6]
Validation set indices: [7 8 9]
Iteration 2
Training set indices: [0 1 2 3 4 5 6 7]
Validation set indices: [8 9]
Iteration 3
Training set indices: [0 1 2 3 4 5 6 7 8]
Validation set indices: [9]

Source code in timecave/validation_methods/OOS.py

def split(self) -> Generator[tuple[np.ndarray, np.ndarray, float], None, None]:
    """
    Split the time series into training and validation sets.

    This method splits the series' indices into disjoint sets containing the training and validation indices.
    At every iteration, an array of training indices and another one containing the validation indices are generated.
    Note that this method is a generator. To access the indices, use the `next()` method or a `for` loop.

    Yields
    ------
    np.ndarray
        Array of training indices.

    np.ndarray
        Array of validation indices.

    float
        Used for compatibility reasons. Irrelevant for this method.

    Examples
    --------
    >>> import numpy as np
    >>> from timecave.validation_methods.OOS import RollingOriginRecalibration
    >>> ts = np.ones(10);
    >>> splitter = RollingOriginRecalibration(ts);
    >>> for ind, (train, val, _) in enumerate(splitter.split()):
    ...
    ...     print(f"Iteration {ind+1}");
    ...     print(f"Training set indices: {train}");
    ...     print(f"Validation set indices: {val}");
    Iteration 1
    Training set indices: [0 1 2 3 4 5 6]
    Validation set indices: [7 8 9]
    Iteration 2
    Training set indices: [0 1 2 3 4 5 6 7]
    Validation set indices: [8 9]
    Iteration 3
    Training set indices: [0 1 2 3 4 5 6 7 8]
    Validation set indices: [9]
    """

    for ind in self._splitting_ind:

        training = self._indices[:ind]
        validation = self._indices[ind:]

        yield (training, validation, 1.0)

statistics()

Compute relevant statistics for both training and validation sets.

This method computes relevant time series features, such as mean, strength-of-trend, etc. for both the whole time series, the training set and the validation set. It can and should be used to ensure that the characteristics of both the training and validation sets are, statistically speaking, similar to those of the time series one wishes to forecast. If this is not the case, using the validation method will most likely lead to a poor assessment of the model's performance.

Returns:

Type	Description
DataFrame	Relevant features for the entire time series.
DataFrame	Relevant features for the training set.
DataFrame	Relevant features for the validation set.

Raises:

Type	Description
ValueError	If the time series is composed of less than three samples.

Examples:

>>> import numpy as np
>>> from timecave.validation_methods.OOS import RollingOriginRecalibration
>>> ts = np.hstack((np.ones(5), np.zeros(5)));
>>> splitter = RollingOriginRecalibration(ts);
>>> ts_stats, training_stats, validation_stats = splitter.statistics();
Frequency features are only meaningful if the correct sampling frequency is passed to the class.
Training and validation set features can only computed if each set is composed of two or more samples.
>>> ts_stats
   Mean  Median  Min  Max  Variance  P2P_amplitude  Trend_slope  Spectral_centroid  Spectral_rolloff  Spectral_entropy  Strength_of_trend  Mean_crossing_rate  Median_crossing_rate
0   0.5     0.5  0.0  1.0      0.25            1.0    -0.151515           0.114058               0.5           0.38717            1.59099            0.111111              0.111111
>>> training_stats
       Mean  Median  Min  Max  Variance  P2P_amplitude  Trend_slope  Spectral_centroid  Spectral_rolloff  Spectral_entropy  Strength_of_trend  Mean_crossing_rate  Median_crossing_rate
0  0.714286     1.0  0.0  1.0  0.204082            1.0    -0.178571           0.094706          0.428571          0.556506           1.212183            0.166667              0.166667
0  0.625000     1.0  0.0  1.0  0.234375            1.0    -0.178571           0.122818          0.500000          0.600876           1.383496            0.142857              0.142857
0  0.555556     1.0  0.0  1.0  0.246914            1.0    -0.166667           0.105483          0.444444          0.385860           1.502496            0.125000              0.125000
>>> validation_stats
   Mean  Median  Min  Max  Variance  P2P_amplitude  Trend_slope  Spectral_centroid  Spectral_rolloff  Spectral_entropy  Strength_of_trend  Mean_crossing_rate  Median_crossing_rate
0   0.0     0.0  0.0  0.0       0.0            0.0          0.0                  0               0.0               0.0                inf                 0.0                   0.0
0   0.0     0.0  0.0  0.0       0.0            0.0          0.0                  0               0.0               0.0                inf                 0.0                   0.0

Source code in timecave/validation_methods/OOS.py

def statistics(self) -> tuple[pd.DataFrame]:
    """
    Compute relevant statistics for both training and validation sets.

    This method computes relevant time series features, such as mean, strength-of-trend, etc. for both the whole time series, the training set and the validation set.
    It can and should be used to ensure that the characteristics of both the training and validation sets are, statistically speaking, similar to those of the time series one wishes to forecast.
    If this is not the case, using the validation method will most likely lead to a poor assessment of the model's performance.

    Returns
    -------
    pd.DataFrame
        Relevant features for the entire time series.

    pd.DataFrame
        Relevant features for the training set.

    pd.DataFrame
        Relevant features for the validation set.

    Raises
    ------
    ValueError
        If the time series is composed of less than three samples.

    Examples
    --------
    >>> import numpy as np
    >>> from timecave.validation_methods.OOS import RollingOriginRecalibration
    >>> ts = np.hstack((np.ones(5), np.zeros(5)));
    >>> splitter = RollingOriginRecalibration(ts);
    >>> ts_stats, training_stats, validation_stats = splitter.statistics();
    Frequency features are only meaningful if the correct sampling frequency is passed to the class.
    Training and validation set features can only computed if each set is composed of two or more samples.
    >>> ts_stats
       Mean  Median  Min  Max  Variance  P2P_amplitude  Trend_slope  Spectral_centroid  Spectral_rolloff  Spectral_entropy  Strength_of_trend  Mean_crossing_rate  Median_crossing_rate
    0   0.5     0.5  0.0  1.0      0.25            1.0    -0.151515           0.114058               0.5           0.38717            1.59099            0.111111              0.111111
    >>> training_stats
           Mean  Median  Min  Max  Variance  P2P_amplitude  Trend_slope  Spectral_centroid  Spectral_rolloff  Spectral_entropy  Strength_of_trend  Mean_crossing_rate  Median_crossing_rate
    0  0.714286     1.0  0.0  1.0  0.204082            1.0    -0.178571           0.094706          0.428571          0.556506           1.212183            0.166667              0.166667
    0  0.625000     1.0  0.0  1.0  0.234375            1.0    -0.178571           0.122818          0.500000          0.600876           1.383496            0.142857              0.142857
    0  0.555556     1.0  0.0  1.0  0.246914            1.0    -0.166667           0.105483          0.444444          0.385860           1.502496            0.125000              0.125000
    >>> validation_stats
       Mean  Median  Min  Max  Variance  P2P_amplitude  Trend_slope  Spectral_centroid  Spectral_rolloff  Spectral_entropy  Strength_of_trend  Mean_crossing_rate  Median_crossing_rate
    0   0.0     0.0  0.0  0.0       0.0            0.0          0.0                  0               0.0               0.0                inf                 0.0                   0.0
    0   0.0     0.0  0.0  0.0       0.0            0.0          0.0                  0               0.0               0.0                inf                 0.0                   0.0
    """

    if self._n_samples <= 2:

        raise ValueError(
            "Basic statistics can only be computed if the time series comprises more than two samples."
        )

    print("Frequency features are only meaningful if the correct sampling frequency is passed to the class.")

    full_features = get_features(self._series, self.sampling_freq)
    training_stats = []
    validation_stats = []

    print(
        "Training and validation set features can only computed if each set is composed of two or more samples."
    )

    for training, validation, _ in self.split():

        if self._series[training].shape[0] >= 2:

            training_feat = get_features(self._series[training], self.sampling_freq)
            training_stats.append(training_feat)

        if self._series[validation].shape[0] >= 2:

            validation_feat = get_features(
                self._series[validation], self.sampling_freq
            )
            validation_stats.append(validation_feat)

    training_features = pd.concat(training_stats)
    validation_features = pd.concat(validation_stats)

    return (full_features, training_features, validation_features)