Source code for evalml.demos.diabetes

"""Load the diabetes dataset, which can be used for regression problems."""

import pandas as pd
import woodwork as ww
from sklearn.preprocessing import scale

import evalml
from evalml.preprocessing import load_data


def load_diabetes():
    """Load diabetes dataset. Used for regression problems.

    Returns:
        (pd.Dataframe, pd.Series): X and y
    """
    filename = (
        "https://oss.alteryx.com/datasets/diabetes-2022-06-27.csv?library=evalml&version="
        + evalml.__version__
    )
    X, y = load_data(filename, index=None, target="target")
    y.name = None

    # This scales the feature variables by the standard deviation times the square root of n_samples
    # This change is necessary due to https://github.com/scikit-learn/scikit-learn/pull/16605
    # In previous versions the diabetes.csv data was returned, but now scikit-learn scales the data then returns
    y = y.astype(float)
    X_np = scale(X.to_numpy(float), copy=False)
    X_np /= X_np.shape[0] ** 0.5
    X = pd.DataFrame(X_np, columns=X.columns)
    X.ww.init()
    y = ww.init_series(y)

    return X, y