anomaly_rolling#

pyCGM2.Signal.anomaly.anomaly_rolling(values: ndarray, aprioriError: int = 0, window: int = 10, threshold: int = 3, method: str = 'median', plot: bool = False, label: str = 'Unknow', referenceValues: float | None = None)#

Detects anomalies in a sequence of values using a rolling window approach.

This function identifies points in the data that are statistically different from the values in a moving window. Anomalies are determined based on a threshold applied to a specified descriptive statistic (median or mean) over a rolling window.

Parameters:

  • values (np.ndarray) – The array of values for anomaly detection.

  • aprioriError (int, optional) – An a priori error margin added and subtracted from each value. Defaults to 0.

  • window (int, optional) – The size of the rolling window to compute the descriptive statistic. Defaults to 10.

  • threshold (int, optional) – The threshold for identifying anomalies, based on the standard deviation factor. Defaults to 3.

  • method (str, optional) – The method of computing the descriptive statistic (‘median’ or ‘mean’). Defaults to “median”.

  • plot (bool, optional) – If True, a plot of the values, rolling statistic, and anomalies is displayed. Defaults to False.

  • label (str, optional) – A label for the data, used in plotting. Defaults to “Unknown”.

  • referenceValues (Optional[float], optional) – An array of reference values to be used instead of the values computed from the rolling window. If None, the rolling window values are used. Defaults to None.

Returns:

list – A list of indices where anomalies are detected in the input values.

Note

  • Zero values in the input are considered anomalies and are warned about.

  • The function plots the original values, the rolling statistic, and the detected anomalies if plot is True.