Source code for evalml.pipelines.components.transformers.dimensionality_reduction.pca

import pandas as pd
from sklearn.decomposition import PCA as SkPCA
from skopt.space import Real

from evalml.pipelines.components.transformers import Transformer
from evalml.utils import (
    _retain_custom_types_and_initalize_woodwork,
    infer_feature_types,
    is_all_numeric,
)


class PCA(Transformer):
    """
    Reduces the number of features by using Principal Component Analysis (PCA).

    Arguments:
        variance (float): The percentage of the original data variance that should be preserved when reducing the
            number of features. Defaults to 0.95.
        n_components (int): The number of features to maintain after computing SVD. Defaults to None, but will override
            variance variable if set.
        random_seed (int): Seed for the random number generator. Defaults to 0.
    """

    name = "PCA Transformer"
    hyperparameter_ranges = {"variance": Real(0.25, 1)}
    """{"variance": Real(0.25, 1)}"""

    def __init__(self, variance=0.95, n_components=None, random_seed=0, **kwargs):
        parameters = {"variance": variance, "n_components": n_components}
        parameters.update(kwargs)
        if n_components:
            pca = SkPCA(n_components=n_components, random_state=random_seed, **kwargs)
        else:
            pca = SkPCA(n_components=variance, random_state=random_seed, **kwargs)
        super().__init__(
            parameters=parameters, component_obj=pca, random_seed=random_seed
        )

    def fit(self, X, y=None):
        X = infer_feature_types(X)
        if not is_all_numeric(X):
            raise ValueError("PCA input must be all numeric")
        self._component_obj.fit(X)
        return self

    def transform(self, X, y=None):
        X_ww = infer_feature_types(X)
        if not is_all_numeric(X_ww):
            raise ValueError("PCA input must be all numeric")
        X_t = self._component_obj.transform(X)
        X_t = pd.DataFrame(
            X_t,
            index=X_ww.index,
            columns=[f"component_{i}" for i in range(X_t.shape[1])],
        )
        return _retain_custom_types_and_initalize_woodwork(X_ww, X_t)

    def fit_transform(self, X, y=None):
        X_ww = infer_feature_types(X)
        if not is_all_numeric(X_ww):
            raise ValueError("PCA input must be all numeric")
        X_t = self._component_obj.fit_transform(X, y)
        X_t = pd.DataFrame(
            X_t,
            index=X_ww.index,
            columns=[f"component_{i}" for i in range(X_t.shape[1])],
        )
        return _retain_custom_types_and_initalize_woodwork(X_ww, X_t)