stream-learn 0.8.4

Python package equipped with a procedures to process data streams using estimators with API compatible with scikit-learn.

stream-learn

stream-learn is a Python package equipped with a procedures to process data streams using estimators with API compatible with scikit-learn.

Documentation

API documentation with set of examples may be found on the documentation page.

Installation

stream-learn is available on the PyPi and you may install it with pip:

pip install stream-learn

Example usage

import strlearn as sl
from sklearn.neural_network import MLPClassifier
from sklearn.naive_bayes import GaussianNB

# Initialize list of scikit-learn classifiers with partial_fit() function
clf = [MLPClassifier(), GaussianNB()]

# Declare data stream
stream = sl.streams.StreamGenerator(n_chunks=10, n_drifts=1)

# Select vector of metrics
metrics = [sl.utils.metrics.bac, sl.utils.metrics.f_score]

# Initialize evaluator with given metrics
evaluator = sl.evaluators.TestThenTrain(metrics)

# Run evaluator over stream with classifier
evaluator.process(stream, clf)

>>> print(evaluator.scores)
[[[0.29730274 0.29145729]
  [0.34494021 0.36097561]
  [0.43464118 0.44878049]
  [0.42579578 0.36666667]
  [0.45569557 0.4171123 ]
  [0.47020869 0.44791667]
  [0.4645207  0.46534653]
  [0.525      0.5177665 ]
  [0.4893617  0.46875   ]]

 [[0.87701288 0.88038278]
  [0.90091448 0.9047619 ]
  [0.89930938 0.9047619 ]
  [0.85376189 0.82681564]
  [0.61521152 0.60913706]
  [0.64714185 0.61538462]
  [0.64556129 0.62564103]
  [0.74       0.74      ]
  [0.80820955 0.80597015]]]