Description

In lines 300-310 of ensemble/partial_dependence, an error message is incorrectly specified:

try:
    for fxs in features:
        l = []
        # explicit loop so "i" is bound for exception below
        for i in fxs:
            l.append(feature_names[i])
        names.append(l)
except IndexError:
    raise ValueError('features[i] must be in [0, n_features) '
                     'but was %d' % i)

Note that the index error can only be thrown by feature_names[i], not features[i] as the error claims. Initially, I expected the feature_names array to correspond to thefeatures array (in particular, that they'd be the same length). Since I was only interested in partial dependence plots for 10 features (out of 60 total), both of the arrays I passed had length 10. I ran into the above error because feature_names did not have length 60. The wording of the error caused me to look at features[i] , which did not help resolve the issue.

The error should clearly indicate that the index error is caused by feature_names[i] to save debugging time.

Steps/Code to Reproduce

You can run the following code to see the error:

import numpy as np
from sklearn.ensemble.partial_dependence import plot_partial_dependence
from sklearn.ensemble import GradientBoostingClassifier

X = np.arange(0,24,1).reshape((8,3))
y = np.ones(8)
model = GradientBoostingClassifier()
y[0] = 0
model.fit(X,y)
features = [1,2]
feature_names = ['IQ', 'Athletic Ability']

pdp_fig = plot_partial_dependence(model, X, features = features, feature_names = feature_names)

Again, note that the problem is caused by the length of the feature_names array, not the features array.

Expected Results

I'd expect the error to correctly indicate which parameter is causing the index error.

Actual Results

Instead, the error reads features[i] must be in [0, n_features) but was 2. In this case n_features == X.shape[1] which is 3. The error claims that 2 is not in [0,3) , making users question their sanity.

Versions

Darwin-15.5.0-x86_64-i386-64bit
('Python', '2.7.12 |Anaconda custom (x86_64)| (default, Jul 2 2016, 17:43:17) \n[GCC 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2336.11.00)]')
('NumPy', '1.11.1')
('SciPy', '0.17.1')
('Scikit-Learn', '0.17.1')