bootstrap_simple.py

from typing import List
import numpy as np
import pandas as pd
from sklearn.base import BaseEstimator

def bootstrap_fit(X: np.ndarray, y: np.array, model_class, B=200, **model_params) -> List[BaseEstimator]:
    """
    Fits and returns a list of B models based upon bootstrap sampling with replacement
    :param X: features to fit model on
    :param y: target (continuous for regression
    :param model_class: any valid sklearn supervised learning class
    :param B: number of bootstrapped models to aggregate
    :param model_params: additional sklearn model parameters, used when instantiating model_class
    :return: a list of fitted sklearn models avaliable to call predict() upon
    """
    assert isinstance(model_class(), BaseEstimator)
    # we need to recombine X and y since we'll need to bootstrap sample from them with replacement
    combined_df = pd.DataFrame(np.column_stack((X, y)))

    bootstrapped_models = []  # create a list to store the fitted models

    for b in range(B):  # bootstrap sample and fit model B times
        # model_class can be any scikit-learn supervised model that subscribes to the fit() and predict() API
        bootstrap_model = model_class(**model_params)
        # sample with replacement
        bootstrap_dataset = combined_df.sample(len(combined_df), replace=True)
        bootstrap_X = bootstrap_dataset.iloc[:, :-1]
        bootstrap_y = bootstrap_dataset.iloc[:, -1]
        # fit model and add it to the collection of bootstrapped models
        bootstrap_model.fit(bootstrap_X, bootstrap_y)
        bootstrapped_models.append(bootstrap_model)
    return bootstrapped_models