AnomalySine

Simulate a stream with anomalies in sine waves.

The amount of data generated by this generator is finite.

The data generated corresponds to sine and cosine functions. Anomalies are induced by replacing the cosine values with values from a different a sine function. The contextual flag can be used to introduce contextual anomalies which are values in the normal global range, but abnormal compared to the seasonal pattern. Contextual attributes are introduced by replacing cosine entries with sine values.

The target indicates whether or not the instances are anomalous.

Parameters

• n_samples

  Type → int

  Default → 10000

  The number of samples to generate. This generator creates a batch of data affected by contextual anomalies and noise.

• n_anomalies

  Type → int

  Default → 2500

  Number of anomalies. Can't be larger than n_samples.

• contextual

  Type → bool

  Default → False

  If True, will add contextual anomalies.

• n_contextual

  Type → int

  Default → 2500

  Number of contextual anomalies. Can't be larger than n_samples.

• shift

  Type → int

  Default → 4

  Shift in number of samples applied when retrieving contextual anomalies.

• noise

  Type → float

  Default → 0.5

  Amount of noise.

• replace

  Type → bool

  Default → True

  If True, anomalies are randomly sampled with replacement.

• seed

  Type → int | None

  Default → None

  Random seed for reproducibility.

Attributes

• desc

  Return the description from the docstring.

Examples

from river.datasets import synth

dataset = synth.AnomalySine(
    seed=12345,
    n_samples=100,
    n_anomalies=25,
    contextual=True,
    n_contextual=10
)

for x, y in dataset.take(5):
    print(x, y)

{'sine': -0.7119, 'cosine': 0.8777} False
{'sine': 0.8792, 'cosine': -0.0290} False
{'sine': 0.0440, 'cosine': 3.0852} True
{'sine': 0.5520, 'cosine': 3.4515} True
{'sine': 0.8037, 'cosine': 0.4027} False

Methods

take

Iterate over the k samples.

Parameters

• k — 'int'