bootstrap_predictions

from sklearn.ensemble import BaggingClassifier

def bootstrap_predictions(estimator, X, y, X_test, n_bootstrap=101):
    """Bootstrap a given classifier.
    
    Parameters
    ----------
    estimator : object
        A classifier instance with sklearn-compatible interface.

    X : array, shape = (n_samples, n_features)
        The X part of the full dataset.

    y : array, shape = (n_samples,)
        The target labels of the full dataset.

    X_test : array, shape(n_test_samples, n_features)
        The test data to measure theoutput of bootstrapped estimators on.

    n_bootstrap : int, nonnegative
        The number of bootstrap replication of `estimator` to make.

    Returns
    -------
    proba : array, shape=(n_test_samples, n_bootstrap), dtype=float
        The matrix of bootstrapped outputs of the classifier.

    bag : list
        The list of bootstrap replications of the classifier.

    Details
    -------
    The `(X, y)` full dataset is used to generate bootstrap samples. Each one
    of `n_bootstrap` samples is used to train a separate copy of the provided
    `estimator`. Each classifier form resulting set of bootstrapped estimators
    is applied to `X_test` and the output is recorded in `proba` array.
    """

    ### BEGIN Solution
    model = BaggingClassifier(base_estimator=estimator, 
                              n_estimators=n_bootstrap, 
                              bootstrap = True,
                              n_jobs = -1,
                              random_state=42)
    
    model.fit(X, y)

    # The list of bootstrap replications of the classifier.
    bag = model.estimators_
    # The matrix of bootstrapped outputs of the classifier.
    proba = np.zeros(shape=(X_test.shape[0], n_bootstrap))

    for index, estimator in enumerate(model.estimators_):
        proba[:, index] = estimator.predict(X_test)


    ### END Solution

    return proba, bag