Alpha Investing¶

skfeature.function.streaming.alpha_investing

Description¶

Alpha Investing is a streamwise (online) feature selection method. Features arrive one at a time and are accepted or rejected on the fly using the alpha-investing rule with a wealth parameter, suitable for binary and univariate regression problems.

Usage¶

import numpy as np
from sklearn.datasets import make_classification
from sklearn.pipeline import Pipeline
from sklearn.svm import SVC

from skfeature.function.streaming import alpha_investing

X, y = make_classification(n_samples=200, n_features=20, n_informative=5)
y = y.astype(float)

pipeline = Pipeline(
    [
        ("alphainvesting", alpha_investing.AlphaInvesting(w0=0.05, dw=0.05)),
        ("svm", SVC()),
    ]
)
pipeline.fit(X, y)

Parameters¶

X: numpy array, shape (n_samples, n_features) — input data, one feature per time step
y: numpy array, shape (n_samples,) — class labels or regression target
w0: float — initial wealth
dw: float — wealth increment

Returns¶

The module also exposes the AlphaInvesting class, a sklearn.base.TransformerMixin for use inside scikit-learn pipelines.

References¶

Zhou, Jing, Foster, Dean P., Stine, Robert A., and Ungar, Lyle H. "Streamwise feature selection." JMLR 2006.