holdenk commented on a change in pull request #23157: [SPARK-26185][PYTHON]add
weightCol in python MulticlassClassificationEvaluator
URL: https://github.com/apache/spark/pull/23157#discussion_r255169107
##########
File path: python/pyspark/mllib/evaluation.py
##########
@@ -192,16 +192,48 @@ class MulticlassMetrics(JavaModelWrapper):
0.66...
>>> metrics.weightedFMeasure(2.0)
0.65...
+ >>> predAndLabelsWithOptWeight = sc.parallelize([(0.0, 0.0, 1.0), (0.0,
1.0, 1.0),
+ ... (0.0, 0.0, 1.0), (1.0, 0.0, 1.0), (1.0, 1.0, 1.0), (1.0, 1.0,
1.0), (1.0, 1.0, 1.0),
+ ... (2.0, 2.0, 1.0), (2.0, 0.0, 1.0)])
+ >>> metrics = MulticlassMetrics(predAndLabelsWithOptWeight)
+ >>> metrics.confusionMatrix().toArray()
+ array([[ 2., 1., 1.],
+ [ 1., 3., 0.],
+ [ 0., 0., 1.]])
+ >>> metrics.falsePositiveRate(0.0)
+ 0.2...
+ >>> metrics.precision(1.0)
+ 0.75...
+ >>> metrics.recall(2.0)
+ 1.0...
+ >>> metrics.fMeasure(0.0, 2.0)
+ 0.52...
+ >>> metrics.accuracy
+ 0.66...
+ >>> metrics.weightedFalsePositiveRate
+ 0.19...
+ >>> metrics.weightedPrecision
+ 0.68...
+ >>> metrics.weightedRecall
+ 0.66...
+ >>> metrics.weightedFMeasure()
+ 0.66...
+ >>> metrics.weightedFMeasure(2.0)
+ 0.65...
.. versionadded:: 1.4.0
"""
- def __init__(self, predictionAndLabels):
- sc = predictionAndLabels.ctx
+ def __init__(self, predAndLabelsWithOptWeight):
Review comment:
My one concern here would be if a user had been using a named parameter to
construct this we would not be backwards compatible, however since the next
release is slated to be Spark 3 I think this relatively minor break is OK.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
