API Reference / datasets / generators /

classification/

Synthetic classification datasets.

Labelled data with a known decision boundary, so you can check what an algorithm recovers. Several also generate concept drift — the boundary moving partway through the stream — which is what streaming learners have to detect and adapt to.

Generators

  • Agrawal: Loan applications over nine attributes, with ten selectable
target functions. A standard stream-mining benchmark; switching function mid-stream produces abrupt drift.
  • Hyperplane: Points labelled by which side of a rotating hyperplane they
fall on. Rotating it slowly gives gradual drift.
  • RandomRBF: Radial basis clusters with random centroids, optionally
drifting.
  • LED: The seven-segment digits, with a tunable proportion of segments
flipped — a controlled way to vary label noise.
python 
>>> from tuiml.datasets.generators.classification import Agrawal
>>> data = Agrawal(n_samples=1000, function=1, random_state=0).generate()
>>> data.X.shape
(1000, 9)

Modules

agrawal

Agrawal generator: synthetic loan-applicant records....

1 class · 0 functions
hyperplane

Hyperplane generator: points split by a random linear boundary....

1 class · 0 functions
led

LED generator: seven-segment digit recognition....

1 class · 0 functions
random_rbf

RandomRBF generator: Gaussian clouds around random centroids....

1 class · 0 functions