Skip to content

[SPARK-23161][PYSPARK][ML]Add missing APIs to Python GBTClassifier #21413

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 5 commits into from
Closed

[SPARK-23161][PYSPARK][ML]Add missing APIs to Python GBTClassifier #21413

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
16d19f4
Add missing APIs to Python GBTClassifier
huaxingao May 23, 2018
0631f8d
address comment
huaxingao May 24, 2018
d60823f
change the default value of featureSubsetStrategy to "all" in GBTClas…
huaxingao May 24, 2018
714ab33
change the deprecation version to 2.4.0
huaxingao May 29, 2018
d8f3906
fix the deprecation version
huaxingao May 30, 2018
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 29 additions & 6 deletions python/pyspark/ml/classification.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1131,6 +1131,13 @@ def setParams(self, featuresCol="features", labelCol="label", predictionCol="pre
def _create_model(self, java_model):
return RandomForestClassificationModel(java_model)

@since("2.4.0")
def setFeatureSubsetStrategy(self, value):
"""
Sets the value of :py:attr:`featureSubsetStrategy`.
"""
return self._set(featureSubsetStrategy=value)


class RandomForestClassificationModel(TreeEnsembleModel, JavaClassificationModel, JavaMLWritable,
JavaMLReadable):
Expand Down Expand Up @@ -1193,6 +1200,8 @@ class GBTClassifier(JavaEstimator, HasFeaturesCol, HasLabelCol, HasPredictionCol
>>> si_model = stringIndexer.fit(df)
>>> td = si_model.transform(df)
>>> gbt = GBTClassifier(maxIter=5, maxDepth=2, labelCol="indexed", seed=42)
>>> gbt.getFeatureSubsetStrategy()
'all'
>>> model = gbt.fit(td)
>>> model.featureImportances
SparseVector(1, {0: 1.0})
Expand Down Expand Up @@ -1226,6 +1235,8 @@ class GBTClassifier(JavaEstimator, HasFeaturesCol, HasLabelCol, HasPredictionCol
... ["indexed", "features"])
>>> model.evaluateEachIteration(validation)
[0.25..., 0.23..., 0.21..., 0.19..., 0.18...]
>>> model.numClasses
2

.. versionadded:: 1.4.0
"""
Expand All @@ -1244,19 +1255,22 @@ class GBTClassifier(JavaEstimator, HasFeaturesCol, HasLabelCol, HasPredictionCol
def __init__(self, featuresCol="features", labelCol="label", predictionCol="prediction",
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,
maxMemoryInMB=256, cacheNodeIds=False, checkpointInterval=10, lossType="logistic",
maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0):
maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0,
featureSubsetStrategy="all"):
"""
__init__(self, featuresCol="features", labelCol="label", predictionCol="prediction", \
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0, \
maxMemoryInMB=256, cacheNodeIds=False, checkpointInterval=10, \
lossType="logistic", maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0)
lossType="logistic", maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0, \
featureSubsetStrategy="all")
"""
super(GBTClassifier, self).__init__()
self._java_obj = self._new_java_obj(
"org.apache.spark.ml.classification.GBTClassifier", self.uid)
self._setDefault(maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,
maxMemoryInMB=256, cacheNodeIds=False, checkpointInterval=10,
lossType="logistic", maxIter=20, stepSize=0.1, subsamplingRate=1.0)
lossType="logistic", maxIter=20, stepSize=0.1, subsamplingRate=1.0,
featureSubsetStrategy="all")
kwargs = self._input_kwargs
self.setParams(**kwargs)

Expand All @@ -1265,12 +1279,14 @@ def __init__(self, featuresCol="features", labelCol="label", predictionCol="pred
def setParams(self, featuresCol="features", labelCol="label", predictionCol="prediction",
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,
maxMemoryInMB=256, cacheNodeIds=False, checkpointInterval=10,
lossType="logistic", maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0):
lossType="logistic", maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0,
featureSubsetStrategy="all"):
"""
setParams(self, featuresCol="features", labelCol="label", predictionCol="prediction", \
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0, \
maxMemoryInMB=256, cacheNodeIds=False, checkpointInterval=10, \
lossType="logistic", maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0)
lossType="logistic", maxIter=20, stepSize=0.1, seed=None, subsamplingRate=1.0, \
featureSubsetStrategy="all")
Sets params for Gradient Boosted Tree Classification.
"""
kwargs = self._input_kwargs
Expand All @@ -1293,8 +1309,15 @@ def getLossType(self):
"""
return self.getOrDefault(self.lossType)

@since("2.4.0")
def setFeatureSubsetStrategy(self, value):
"""
Sets the value of :py:attr:`featureSubsetStrategy`.
"""
return self._set(featureSubsetStrategy=value)


class GBTClassificationModel(TreeEnsembleModel, JavaPredictionModel, JavaMLWritable,
class GBTClassificationModel(TreeEnsembleModel, JavaClassificationModel, JavaMLWritable,
JavaMLReadable):
"""
Model fitted by GBTClassifier.
Expand Down
70 changes: 45 additions & 25 deletions python/pyspark/ml/regression.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -602,6 +602,14 @@ class TreeEnsembleParams(DecisionTreeParams):
"used for learning each decision tree, in range (0, 1].",
typeConverter=TypeConverters.toFloat)

supportedFeatureSubsetStrategies = ["auto", "all", "onethird", "sqrt", "log2"]

featureSubsetStrategy = \
Param(Params._dummy(), "featureSubsetStrategy",
"The number of features to consider for splits at each tree node. Supported " +
"options: " + ", ".join(supportedFeatureSubsetStrategies) + ", (0.0-1.0], [1-n].",
typeConverter=TypeConverters.toString)

def __init__(self):
super(TreeEnsembleParams, self).__init__()

Expand All @@ -619,6 +627,22 @@ def getSubsamplingRate(self):
"""
return self.getOrDefault(self.subsamplingRate)

@since("1.4.0")
def setFeatureSubsetStrategy(self, value):
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this method should be in the GBT/RandomForest estimator classes (classification and regression) and the old method in RandomForestParams should be deprecated, would you mind doing this?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@BryanCutler Thanks for your review. I will modify the code. One question: Shall I only put the setFeatureSubsetStrategy in GBT/RandomForest? or put both the setter and getter? I looked the ml python code, it seems to me that the getter and setter are always in pairs.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

setFeatureSubsetStrategy should only be in the GBT/RF estimators, while getFeatureSubsetStrategy can be in TreeEnsembleParams so it is inherited by both the estimators and models. It's because we don't want methods to set training params in the Model classes.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Got it. Thanks a lot!

"""
Sets the value of :py:attr:`featureSubsetStrategy`.

.. note:: Deprecated in 2.4.0 and will be removed in 3.0.0.
"""
return self._set(featureSubsetStrategy=value)

@since("1.4.0")
def getFeatureSubsetStrategy(self):
"""
Gets the value of featureSubsetStrategy or its default value.
"""
return self.getOrDefault(self.featureSubsetStrategy)


class TreeRegressorParams(Params):
"""
Expand Down Expand Up @@ -654,14 +678,8 @@ class RandomForestParams(TreeEnsembleParams):
Private class to track supported random forest parameters.
"""

supportedFeatureSubsetStrategies = ["auto", "all", "onethird", "sqrt", "log2"]
numTrees = Param(Params._dummy(), "numTrees", "Number of trees to train (>= 1).",
typeConverter=TypeConverters.toInt)
featureSubsetStrategy = \
Param(Params._dummy(), "featureSubsetStrategy",
"The number of features to consider for splits at each tree node. Supported " +
"options: " + ", ".join(supportedFeatureSubsetStrategies) + ", (0.0-1.0], [1-n].",
typeConverter=TypeConverters.toString)

def __init__(self):
super(RandomForestParams, self).__init__()
Expand All @@ -680,20 +698,6 @@ def getNumTrees(self):
"""
return self.getOrDefault(self.numTrees)

@since("1.4.0")
def setFeatureSubsetStrategy(self, value):
"""
Sets the value of :py:attr:`featureSubsetStrategy`.
"""
return self._set(featureSubsetStrategy=value)

@since("1.4.0")
def getFeatureSubsetStrategy(self):
"""
Gets the value of featureSubsetStrategy or its default value.
"""
return self.getOrDefault(self.featureSubsetStrategy)


class GBTParams(TreeEnsembleParams):
"""
Expand Down Expand Up @@ -981,6 +985,13 @@ def setParams(self, featuresCol="features", labelCol="label", predictionCol="pre
def _create_model(self, java_model):
return RandomForestRegressionModel(java_model)

@since("2.4.0")
def setFeatureSubsetStrategy(self, value):
"""
Sets the value of :py:attr:`featureSubsetStrategy`.
"""
return self._set(featureSubsetStrategy=value)


class RandomForestRegressionModel(TreeEnsembleModel, JavaPredictionModel, JavaMLWritable,
JavaMLReadable):
Expand Down Expand Up @@ -1029,6 +1040,8 @@ class GBTRegressor(JavaEstimator, HasFeaturesCol, HasLabelCol, HasPredictionCol,
>>> gbt = GBTRegressor(maxIter=5, maxDepth=2, seed=42)
>>> print(gbt.getImpurity())
variance
>>> print(gbt.getFeatureSubsetStrategy())
all
>>> model = gbt.fit(df)
>>> model.featureImportances
SparseVector(1, {0: 1.0})
Expand Down Expand Up @@ -1079,20 +1092,20 @@ def __init__(self, featuresCol="features", labelCol="label", predictionCol="pred
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,
maxMemoryInMB=256, cacheNodeIds=False, subsamplingRate=1.0,
checkpointInterval=10, lossType="squared", maxIter=20, stepSize=0.1, seed=None,
impurity="variance"):
impurity="variance", featureSubsetStrategy="all"):
"""
__init__(self, featuresCol="features", labelCol="label", predictionCol="prediction", \
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0, \
maxMemoryInMB=256, cacheNodeIds=False, subsamplingRate=1.0, \
checkpointInterval=10, lossType="squared", maxIter=20, stepSize=0.1, seed=None, \
impurity="variance")
impurity="variance", featureSubsetStrategy="all")
"""
super(GBTRegressor, self).__init__()
self._java_obj = self._new_java_obj("org.apache.spark.ml.regression.GBTRegressor", self.uid)
self._setDefault(maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,
maxMemoryInMB=256, cacheNodeIds=False, subsamplingRate=1.0,
checkpointInterval=10, lossType="squared", maxIter=20, stepSize=0.1,
impurity="variance")
impurity="variance", featureSubsetStrategy="all")
kwargs = self._input_kwargs
self.setParams(**kwargs)

Expand All @@ -1102,13 +1115,13 @@ def setParams(self, featuresCol="features", labelCol="label", predictionCol="pre
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0,
maxMemoryInMB=256, cacheNodeIds=False, subsamplingRate=1.0,
checkpointInterval=10, lossType="squared", maxIter=20, stepSize=0.1, seed=None,
impuriy="variance"):
impuriy="variance", featureSubsetStrategy="all"):
"""
setParams(self, featuresCol="features", labelCol="label", predictionCol="prediction", \
maxDepth=5, maxBins=32, minInstancesPerNode=1, minInfoGain=0.0, \
maxMemoryInMB=256, cacheNodeIds=False, subsamplingRate=1.0, \
checkpointInterval=10, lossType="squared", maxIter=20, stepSize=0.1, seed=None, \
impurity="variance")
impurity="variance", featureSubsetStrategy="all")
Sets params for Gradient Boosted Tree Regression.
"""
kwargs = self._input_kwargs
Expand All @@ -1131,6 +1144,13 @@ def getLossType(self):
"""
return self.getOrDefault(self.lossType)

@since("2.4.0")
def setFeatureSubsetStrategy(self, value):
"""
Sets the value of :py:attr:`featureSubsetStrategy`.
"""
return self._set(featureSubsetStrategy=value)


class GBTRegressionModel(TreeEnsembleModel, JavaPredictionModel, JavaMLWritable, JavaMLReadable):
"""
Expand Down