Sktime Classification

Introduction

The sktime classification integration enables hyperparameter optimization of time series classifiers using proper cross‑validation and scoring. Use it via:

• Experiment: SktimeClassificationExperiment
• Estimator interface: hyperactive.integrations.sktime.TSCOptCV

Key Features

• Works with sktime classifiers and pipelines
• Flexible CV strategy and scoring coercion
• Clean separation between optimizer and experiment

Basic Usage

from hyperactive.experiment.integrations import SktimeClassificationExperiment
from hyperactive.opt import RandomSearchSk

# prepare sktime data X, y and a classifier `clf`
exp = SktimeClassificationExperiment(estimator=clf, X=X, y=y, cv=3)
opt = RandomSearchSk(
    param_distributions={"param": ["a", "b"]},
    n_iter=20,
    experiment=exp,
)
best_params = opt.solve()

With TSCOptCV

from hyperactive.integrations.sktime import TSCOptCV
from hyperactive.opt import TPEOptimizer

tsc = TSCOptCV(
    estimator=clf,
    optimizer=TPEOptimizer(param_space={"param": (0.1, 1.0)}, n_trials=30),
    cv=3,
)
tsc.fit(X, y)
print(tsc.best_params_)

Notes

• Choose scoring appropriate for your task (string or callable)
• For forecasting (not classification), use SktimeForecastingExperiment