8000 Merge pull request #4253 from Flimm/doc-precision_recall_fscore_support · scikit-learn/scikit-learn@7ce32d7 · GitHub
[go: up one dir, main page]
Skip to content

Commit 7ce32d7

Browse files
GaelVaroquauxGaelVaroquaux
committed
Merge pull request #4253 from Flimm/doc-precision_recall_fscore_support
[MRG + 1] Add example for precision_recall_fscore_support
2 parents 42229fd + 1250974 commit 7ce32d7

File tree

1 file changed

+12
-2
lines changed

Original file line numberDiff line numberDiff line change
@@ -918,8 +918,8 @@ def precision_recall_fscore_support(y_true, y_pred, beta=1.0, labels=None,
918918
Examples
919919
--------
920920
>>> from sklearn.metrics import precision_recall_fscore_support
921-
>>> y_true = np.array([0, 1, 2, 0, 1, 2])
922-
>>> y_pred = np.array([0, 2, 1, 0, 0, 1])
921+
>>> y_true = np.array(['cat', 'dog', 'pig', 'cat', 'dog', 'pig'])
922+
>>> y_pred = np.array(['cat', 'pig', 'dog', 'cat', 'cat', 'dog'])
923923
>>> precision_recall_fscore_support(y_true, y_pred, average='macro')
924924
... # doctest: +ELLIPSIS
925925
(0.22..., 0.33..., 0.26..., None)
@@ -930,6 +930,16 @@ def precision_recall_fscore_support(y_true, y_pred, beta=1.0, labels=None,
930930
... # doctest: +ELLIPSIS
931931
(0.22..., 0.33..., 0.26..., None)
932932
933+
It is possible to compute per-label precisions, recalls, F1-scores and
934+
supports instead of averaging:
935+
>>> precision_recall_fscore_support(y_true, y_pred, average=None,
936+
... labels=['pig', 'dog', 'cat'])
937+
... # doctest: +ELLIPSIS,+NORMALIZE_WHITESPACE
938+
(array([ 0. , 0. , 0.66...]),
939+
array([ 0., 0., 1.]),
940+
array([ 0. , 0. , 0.8]),
941+
array([2, 2, 2]))
942+
933943
"""
934944
average_options = (None, 'micro', 'macro', 'weighted', 'samples')
935945
if average not in average_options and average != 'binary':