cross_val_regressor.py

# -*- coding: utf-8 -*-

from sklearn.base import BaseEstimator, RegressorMixin, TransformerMixin
from sklearn.cross_validation import cross_val_predict


class ThroughTransformer(BaseEstimator, TransformerMixin):
    def fit(self, X, y=None, **fit_param):
        return self

    def transform(self, X):
        return X


class CrossValRegressor(BaseEstimator, RegressorMixin):
    def __init__(self, estimator):
        self.estimator = estimator

    def fit(self, X, y=None, **fit_param):
        self.estimator.fit(X, y, **fit_param)
        return self

    def transform(self, X):
        return self.predict(X).reshape(-1, 1)

    def fit_transform(self, X, y, **fit_param):
        self.estimator.fit(X, y, **fit_param)
        return cross_val_predict(self.estimator, X, y).reshape(-1, 1)

    def predict(self, X):
        return self.estimator.predict(X)

if __name__ == '__main__':
    from sklearn.datasets import load_diabetes
    from sklearn.cross_validation import train_test_split
    from sklearn.ensemble import RandomForestRegressor, ExtraTreesRegressor
    from sklearn.neighbors import KNeighborsRegressor
    from sklearn.linear_model import LinearRegression, ElasticNet, Lasso
    from sklearn.svm import SVR
    from sklearn.pipeline import Pipeline, FeatureUnion

    diabetes = load_diabetes()
    X_train, X_test, y_train, y_test = train_test_split(diabetes.data, diabetes.target, test_size=0.5, random_state=0)

    features1 = FeatureUnion([
        ('th', ThroughTransformer()),
        ('rf', CrossValRegressor(RandomForestRegressor(random_state=2016))),
        ('et', CrossValRegressor(ExtraTreesRegressor(random_state=2016))),
        ('kn', CrossValRegressor(KNeighborsRegressor())),
        ('svr', CrossValRegressor(SVR())),
    ])
    features2 = FeatureUnion([
        ('rf_10', CrossValRegressor(RandomForestRegressor(n_estimators=10, random_state=0))),
        ('rf_20', CrossValRegressor(RandomForestRegressor(n_estimators=20, random_state=0)))
    ])
    pipe = Pipeline([
        ('f1', features1),
        ('f2', features2),
        ('rf', RandomForestRegressor(random_state=0))
    ])
    pipe.fit(X_train, y_train)
    print(pipe.score(X_test, y_test))

    model = LinearRegression()
    model.fit(X_train, y_train)
    print(model.score(X_test, y_test))

    model = RandomForestRegressor(random_state=2016)
    model.fit(X_train, y_train)
    print(model.score(X_test, y_test))