QuantileFilter

Threshold anomaly filter.

Parameters

• anomaly_detector

  An anomaly detector.

• q

  Type → float

  The quantile level above which to classify an anomaly score as anomalous.

• protect_anomaly_detector

  Default → True

  Indicates whether or not the anomaly detector should be updated when the anomaly score is anomalous. If the data contains sporadic anomalies, then the anomaly detector should likely not be updated. Indeed, if it learns the anomaly score, then it will slowly start to consider anomalous anomaly scores as normal. This might be desirable, for instance in the case of drift.

Attributes

• q

Examples

from river import anomaly
from river import compose
from river import datasets
from river import metrics
from river import preprocessing

model = compose.Pipeline(
    preprocessing.MinMaxScaler(),
    anomaly.QuantileFilter(
        anomaly.HalfSpaceTrees(seed=42),
        q=0.95
    )
)

report = metrics.ClassificationReport()

for x, y in datasets.CreditCard().take(2000):
    score = model.score_one(x)
    is_anomaly = model['QuantileFilter'].classify(score)
    model = model.learn_one(x)
    report = report.update(y, is_anomaly)

report

               Precision   Recall   F1       Support
<BLANKLINE>
       0      99.95%   94.49%   97.14%      1998
       1       0.90%   50.00%    1.77%         2
<BLANKLINE>
   Macro      50.42%   72.25%   49.46%
   Micro      94.45%   94.45%   94.45%
Weighted      99.85%   94.45%   97.05%
<BLANKLINE>
                 94.45% accuracy

Methods

classify

Classify an anomaly score as anomalous or not.

Parameters

• score — 'float'

Returns

bool: A boolean value indicating whether the anomaly score is anomalous or not.

learn_one

Update the anomaly filter and the underlying anomaly detector.

Parameters

• args
• learn_kwargs

Returns

self

score_one

Return an outlier score.

A high score is indicative of an anomaly. A low score corresponds to a normal observation.

Parameters

• args
• kwargs

Returns

An anomaly score. A high score is indicative of an anomaly. A low score corresponds a