RandomSampler

Random sampling by mixing under-sampling and over-sampling.

This is a wrapper for classifiers. It will train the provided classifier by both under-sampling and over-sampling the stream of given observations so that the class distribution seen by the classifier follows a given desired distribution.

See Working with imbalanced data for example usage.

Parameters

• classifier

  Type → base.Classifier

• desired_dist

  Type → dict

  The desired class distribution. The keys are the classes whilst the values are the desired class percentages. The values must sum up to 1. If set to None, then the observations will be sampled uniformly at random, which is strictly equivalent to using ensemble.BaggingClassifier.

• sampling_rate

  Default → 1.0

  The desired ratio of data to sample.

• seed

  Type → int | None

  Default → None

  Random seed for reproducibility.

Examples

from river import datasets
from river import evaluate
from river import imblearn
from river import linear_model
from river import metrics
from river import preprocessing

model = imblearn.RandomSampler(
    (
        preprocessing.StandardScaler() |
        linear_model.LogisticRegression()
    ),
    desired_dist={False: 0.4, True: 0.6},
    sampling_rate=0.8,
    seed=42
)

dataset = datasets.CreditCard().take(3000)

metric = metrics.LogLoss()

evaluate.progressive_val_score(dataset, model, metric)

LogLoss: 0.09...

Methods

learn_one

Update the model with a set of features x and a label y.

Parameters

• x — 'dict[base.typing.FeatureName, Any]'
• y — 'base.typing.ClfTarget'
• kwargs

predict_one

Predict the label of a set of features x.

Parameters

• x
• kwargs

Returns

The predicted label.

predict_proba_one

Predict the probability of each label for a dictionary of features x.

Parameters

• x
• kwargs

Returns

A dictionary that associates a probability which each label.