在RandomForestRegressors列表上调用np.array将返回DecisionTreeRegressors数组

Question

When I try to convert a list of RandomForestRegressors to a numpy array, I get an array of Decision Trees. How do I get an array of RandomForestRegressors instead?

eg

clf0=RandomForestRegressor()
clf1=RandomForestRegressor()

X = np.random.randn(10,1)
y = np.random.randn(10,1)

clf0.fit(X,y)
clf1.fit(X,y)

m = np.array( [clf0, clf1 ] )

m.shape

Out[1]: (2, 10)

I want m.shape to be (2,), not (2,10). And it should contrain RandomForestRegressors, not DecisionTreeRegressors.

Answer 1

This was recently answered on the scikit-learn mailing list: a random forest behaves as sequence of decision trees:

>>> len(clf0)
10
>>> clf0[:2]
[DecisionTreeRegressor(compute_importances=None, criterion='mse',
           max_depth=None, max_features='auto', max_leaf_nodes=None,
           min_density=None, min_samples_leaf=1, min_samples_split=2,
           random_state=1087328618, splitter='best'),
 DecisionTreeRegressor(compute_importances=None, criterion='mse',
           max_depth=None, max_features='auto', max_leaf_nodes=None,
           min_density=None, min_samples_leaf=1, min_samples_split=2,
           random_state=357905606, splitter='best')]

NumPy picks this up and converts the RFs themselves to arrays. The workaround is easy:

>>> m = np.empty(2, object)
>>> m[:] = clf0, clf1
>>> m
array([ RandomForestRegressor(bootstrap=True, compute_importances=None,
           criterion='mse', max_depth=None, max_features='auto',
           max_leaf_nodes=None, min_density=None, min_samples_leaf=1,
           min_samples_split=2, n_estimators=10, n_jobs=1, oob_score=False,
           random_state=None, verbose=0),
       RandomForestRegressor(bootstrap=True, compute_importances=None,
           criterion='mse', max_depth=None, max_features='auto',
           max_leaf_nodes=None, min_density=None, min_samples_leaf=1,
           min_samples_split=2, n_estimators=10, n_jobs=1, oob_score=False,
           random_state=None, verbose=0)], dtype=object)

(Mind you, your y has the wrong shape. It should be (10,) , not (10, 1) .)