from typing import Literal
from pyod.models.ocsvm import OCSVM
from dtaianomaly.anomaly_detection._BaseDetector import Supervision
from dtaianomaly.anomaly_detection._BasePyODAnomalyDetector import (
BasePyODAnomalyDetector,
)
from dtaianomaly.type_validation import LiteralAttribute
from dtaianomaly.windowing import WINDOW_SIZE_TYPE
__all__ = ["OneClassSupportVectorMachine"]
[docs]
class OneClassSupportVectorMachine(BasePyODAnomalyDetector):
"""
Anomaly detector based on One-Class Support Vector Machines (OC-SVM) :cite:`scholkopf1999support`.
The OC-SVM uses a Support Vector Machine to learn
a boundary around the normal behavior with minimal margin. New data can
then be identified as anomaly or not, depending on if the data falls within
this boundary (and thus is normal) or outside the boundary (and thus is
anomalous).
Parameters
----------
window_size : int or str
The window size to use for extracting sliding windows from the time series. This
value will be passed to :py:meth:`~dtaianomaly.anomaly_detection.compute_window_size`.
stride : int, default=1
The stride, i.e., the step size for extracting sliding windows from the time series.
kernel : {'linear', 'poly', 'rbf', 'sigmoid', 'cosine'}, default='rbf'
The kernel to use for PCA.
**kwargs
Arguments to be passed to the PyOD OC-SVM.
Attributes
----------
window_size_ : int
The effectively used window size for this anomaly detector
pyod_detector_ : OCSVM
A OCSVM-detector of PyOD
Examples
--------
>>> from dtaianomaly.anomaly_detection import OneClassSupportVectorMachine
>>> from dtaianomaly.data import demonstration_time_series
>>> x, y = demonstration_time_series()
>>> ocsvm = OneClassSupportVectorMachine(10).fit(x)
>>> ocsvm.decision_function(x) # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
array([-0.7442125 , -1.57019847, -1.86868112, ..., 13.33883568, 12.6492399 , 11.8761641 ]...)
"""
kernel: Literal["linear", "poly", "rbf", "sigmoid", "cosine"]
attribute_validation = {
"kernel": LiteralAttribute("linear", "poly", "rbf", "sigmoid", "cosine")
}
def __init__(
self,
window_size: WINDOW_SIZE_TYPE,
stride: int = 1,
kernel: Literal["linear", "poly", "rbf", "sigmoid", "cosine"] = "rbf",
**kwargs,
):
self.kernel = kernel
super().__init__(window_size, stride, **kwargs)
def _initialize_detector(self, **kwargs) -> OCSVM:
return OCSVM(kernel=self.kernel, **kwargs)
def _supervision(self):
return Supervision.SEMI_SUPERVISED