Source code for pysad.models.median_absolute_deviation

from pysad.core.base_model import BaseModel
from pysad.statistics.median_meter import MedianMeter


class MedianAbsoluteDeviation(BaseModel):
    """Median Absolute Deviation method :cite:`hochenbaum2017automatic`.

    Args:
        absolute (bool): Whether to output score's absolute value (Default=True).
        b (float): The default value `1.4826` is used for normally distributed data. See :cite:`hochenbaum2017automatic` for details. (Default=1.4826).
    """

    def __init__(self, absolute=True, b=1.4826):
        self.b = b
        self.absolute = absolute
        self.median_meter = MedianMeter()
        self.mad_meter = MedianMeter()

    def fit_partial(self, X, y=None):
        """Fits the model to next instance.

        Args:
            X (float): The instance to fit. Note that this model is univariate.
            y (int): Ignored since the model is unsupervised (Default=None).

        Returns:
            object: Returns the self.
        """
        assert len(X) == 1  # Only for time series

        self.median_meter.update(X)
        self.mad_meter.update(X)

        return self

    def score_partial(self, X):
        """Scores the anomalousness of the next instance.

        Args:
            X (float): The instance to score. Higher scores represent more anomalous instances whereas lower scores correspond to more normal instances. Note that this model is univariate.

        Returns:
            float: The anomalousness score of the input instance.
        """
        median = self.median_meter.get()
        mad = self.b * self.mad_meter.get()
        score = (X - median) / (mad + 1e-10)

        return abs(score) if self.absolute else score