MulticlassMetrics#

class pyspark.mllib.evaluation.MulticlassMetrics(predictionAndLabels)[source]#

Evaluator for multiclass classification.

New in version 1.4.0.

Parameters

predictionAndLabelspyspark.RDD: an RDD of prediction, label, optional weight and optional probability.

Examples

>>> predictionAndLabels = sc.parallelize([(0.0, 0.0), (0.0, 1.0), (0.0, 0.0),
...     (1.0, 0.0), (1.0, 1.0), (1.0, 1.0), (1.0, 1.0), (2.0, 2.0), (2.0, 0.0)])
>>> metrics = MulticlassMetrics(predictionAndLabels)
>>> 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...
>>> 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...
>>> predictionAndLabelsWithProbabilities = sc.parallelize([
...      (1.0, 1.0, 1.0, [0.1, 0.8, 0.1]), (0.0, 2.0, 1.0, [0.9, 0.05, 0.05]),
...      (0.0, 0.0, 1.0, [0.8, 0.2, 0.0]), (1.0, 1.0, 1.0, [0.3, 0.65, 0.05])])
>>> metrics = MulticlassMetrics(predictionAndLabelsWithProbabilities)
>>> metrics.logLoss()
0.9682...

Methods

`call`(name, *a)	Call method of java_model
`confusionMatrix`()	Returns confusion matrix: predicted classes are in columns, they are ordered by class label ascending, as in "labels".
`fMeasure`(label[, beta])	Returns f-measure.
`falsePositiveRate`(label)	Returns false positive rate for a given label (category).
`logLoss`([eps])	Returns weighted logLoss.
`precision`(label)	Returns precision.
`recall`(label)	Returns recall.
`truePositiveRate`(label)	Returns true positive rate for a given label (category).
`weightedFMeasure`([beta])	Returns weighted averaged f-measure.

Attributes

`accuracy`	Returns accuracy (equals to the total number of correctly classified instances out of the total number of instances).
`weightedFalsePositiveRate`	Returns weighted false positive rate.
`weightedPrecision`	Returns weighted averaged precision.
`weightedRecall`	Returns weighted averaged recall.
`weightedTruePositiveRate`	Returns weighted true positive rate.

Methods Documentation

call(name, *a)#: Call method of java_model

confusionMatrix()[source]#: Returns confusion matrix: predicted classes are in columns, they are ordered by class label ascending, as in “labels”.

New in version 1.4.0.

fMeasure(label, beta=None)[source]#: Returns f-measure.

New in version 1.4.0.

falsePositiveRate(label)[source]#: Returns false positive rate for a given label (category).

New in version 1.4.0.

logLoss(eps=1e-15)[source]#: Returns weighted logLoss.

New in version 3.0.0.

precision(label)[source]#: Returns precision.

New in version 1.4.0.

recall(label)[source]#: Returns recall.

New in version 1.4.0.

truePositiveRate(label)[source]#: Returns true positive rate for a given label (category).

New in version 1.4.0.

weightedFMeasure(beta=None)[source]#: Returns weighted averaged f-measure.

New in version 1.4.0.

Attributes Documentation

accuracy#: Returns accuracy (equals to the total number of correctly classified instances out of the total number of instances).

New in version 2.0.0.

weightedFalsePositiveRate#: Returns weighted false positive rate.

New in version 1.4.0.

weightedPrecision#: Returns weighted averaged precision.

New in version 1.4.0.

weightedRecall#: Returns weighted averaged recall. (equals to precision, recall and f-measure)

New in version 1.4.0.

weightedTruePositiveRate#: Returns weighted true positive rate. (equals to precision, recall and f-measure)

New in version 1.4.0.