weka.classifiers.sklearn

Class ScikitLearnClassifier

java.lang.Object

weka.classifiers.AbstractClassifier

weka.classifiers.sklearn.ScikitLearnClassifier

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, weka.classifiers.Classifier, weka.core.BatchPredictor, weka.core.CapabilitiesHandler, weka.core.CapabilitiesIgnorer, weka.core.CommandlineRunnable, weka.core.OptionHandler, weka.core.RevisionHandler

public class ScikitLearnClassifier
extends weka.classifiers.AbstractClassifier
implements weka.core.BatchPredictor, weka.core.CapabilitiesHandler

Wrapper classifier for classifiers and regressors implemented in the scikit-learn Python package.

Version:

$Revision: $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ScikitLearnClassifier.Learner Holds info on the different learners available

Field Summary

Fields

Modifier and Type	Field and Description
static weka.core.Tag[]	TAGS_LEARNER The tags for the GUI drop-down for learner selection

Fields inherited from class weka.classifiers.AbstractClassifier

BATCH_SIZE_DEFAULT, NUM_DECIMAL_PLACES_DEFAULT

Constructor Summary

Constructors

Constructor and Description
ScikitLearnClassifier()

Method Summary

Modifier and Type	Method and Description
void	buildClassifier(weka.core.Instances data) Build the classifier
double[]	distributionForInstance(weka.core.Instance instance) Return the predicted probabilities for the supplied instance
double[][]	distributionsForInstances(weka.core.Instances insts) Return the predicted probabilities for the supplied instances
java.lang.String	getBatchSize()
weka.core.Capabilities	getCapabilities() Get the capabilities of this learner
boolean	getContinueOnSysErr() Get whether to try and continue after seeing output on the sys error stream.
boolean	getDontFetchModelFromPython() If true then don't retrieve the model from python.
weka.core.SelectedTag	getLearner() Get the scikit-learn scheme to use
java.lang.String	getLearnerOpts() Get the parameters to pass to the scikit-learn scheme
java.lang.String	getPythonCommand() Get the python command to use.
java.lang.String	getPythonPath() Get optional entries to prepend to the PATH so that python can execute correctly.
java.lang.String	getServerID() Get an optional server name by which to identify the python server to use.
boolean	getUseSupervisedNominalToBinary() Get whether to use the supervised version of nominal to binary
java.lang.String	globalInfo() Global help info
boolean	implementsMoreEfficientBatchPrediction() Returns true, as we send entire test sets over to python for prediction
static void	main(java.lang.String[] argv) Main method for testing this class
void	setBatchSize(java.lang.String size)
void	setContinueOnSysErr(boolean c) Set whether to try and continue after seeing output on the sys error stream.
void	setDontFetchModelFromPython(boolean dontFetchModelFromPython) If true then don't retrieve the model from python.
void	setLearner(weka.core.SelectedTag learner) Set the scikit-learn scheme to use
void	setLearnerOpts(java.lang.String opts) Set the parameters to pass to the scikit-learn scheme
void	setPythonCommand(java.lang.String pyCommand) Set the python command to use.
void	setPythonPath(java.lang.String pythonPath) Set optional entries to prepend to the PATH so that python can execute correctly.
void	setServerID(java.lang.String serverID) Set an optional server name by which to identify the python server to use.
void	setUseSupervisedNominalToBinary(boolean useSupervisedNominalToBinary) Set whether to use the supervised version of nominal to binary
java.lang.String	toString() Get a textual description of this scheme

Methods inherited from class weka.classifiers.AbstractClassifier

batchSizeTipText, classifyInstance, debugTipText, doNotCheckCapabilitiesTipText, forName, getDebug, getDoNotCheckCapabilities, getNumDecimalPlaces, getOptions, getRevision, listOptions, makeCopies, makeCopy, numDecimalPlacesTipText, postExecution, preExecution, run, runClassifier, setDebug, setDoNotCheckCapabilities, setNumDecimalPlaces, setOptions

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, wait, wait, wait

Methods inherited from interface weka.core.OptionHandler

makeCopy

Field Detail

TAGS_LEARNER

public static final weka.core.Tag[] TAGS_LEARNER

The tags for the GUI drop-down for learner selection

Constructor Detail

ScikitLearnClassifier

public ScikitLearnClassifier()

Method Detail

globalInfo

public java.lang.String globalInfo()

Global help info

Returns:

the global help info for this scheme

getCapabilities

public weka.core.Capabilities getCapabilities()

Get the capabilities of this learner

Specified by:

getCapabilities in interface weka.classifiers.Classifier

Specified by:

getCapabilities in interface weka.core.CapabilitiesHandler

Overrides:

getCapabilities in class weka.classifiers.AbstractClassifier

Returns:

the capabilities of this scheme

getUseSupervisedNominalToBinary

@OptionMetadata(displayName="Use supervised nominal to binary conversion",
                description="Use supervised nominal to binary conversion of nominal attributes.",
                commandLineParamName="S",
                commandLineParamSynopsis="-S",
                commandLineParamIsFlag=true,
                displayOrder=3)
public boolean getUseSupervisedNominalToBinary()

Get whether to use the supervised version of nominal to binary

Returns:

true if supervised nominal to binary is to be used

setUseSupervisedNominalToBinary

public void setUseSupervisedNominalToBinary(boolean useSupervisedNominalToBinary)

Set whether to use the supervised version of nominal to binary

Parameters:

useSupervisedNominalToBinary - true if supervised nominal to binary is to be used

getLearner

@OptionMetadata(displayName="Scikit-learn learner",
                description="Scikit-learn learner to use.\nAvailable learners:\nDecisionTreeClassifier, DecisionTreeRegressor, GaussianNB, MultinomialNB,BernoulliNB, LDA, QDA, LogisticRegression, LogisticRegressionCV,\nLinearRegression, ARDRegression, BayesianRidge, ElasticNet, Lars,\nLarsCV, Lasso, LassoCV, LassoLars, LassoLarsCV, LassoLarsIC, MLPClassifier, MLPRegressor, OrthogonalMatchingPursuit,\nOrthogonalMatchingPursuitCV, PassiveAggressiveClassifier, PassiveAggressiveRegressor, Perceptron, RANSACRegressor,\nRidge, RidgeClassifier, RidgeClassifierCV, RidgeCV, SGDClassifier,\nSGDRegressor,TheilSenRegressor, GaussianProcess, KernelRidge, KNeighborsClassifier, \nRadiusNeighborsClassifier, KNeighborsRegressor, RadiusNeighborsRegressor, SVC,\nLinearSVC, NuSVC, SVR, NuSVR, AdaBoostClassifier, AdaBoostRegressor,BaggingClassifier, BaggingRegressor,\nExtraTreeClassifier, ExtraTreeRegressor,GradientBoostingClassifier, GradientBoostingRegressor,\nRandomForestClassifier, RandomForestRegressor, XGBClassifier, XGBRegressor.\n(default = DecisionTreeClassifier)",
                commandLineParamName="learner",
                commandLineParamSynopsis="-learner <learner name>",
                displayOrder=1)
public weka.core.SelectedTag getLearner()

Get the scikit-learn scheme to use

Returns:

the scikit-learn scheme to use

setLearner

public void setLearner(weka.core.SelectedTag learner)

Set the scikit-learn scheme to use

Parameters:

learner - the scikit-learn scheme to use

getLearnerOpts

@OptionMetadata(displayName="Learner parameters",
                description="learner parameters to use",
                displayOrder=2,
                commandLineParamName="parameters",
                commandLineParamSynopsis="-parameters <comma-separated list of name=value pairs>")
public java.lang.String getLearnerOpts()

Get the parameters to pass to the scikit-learn scheme

Returns:

the parameters to use

setLearnerOpts

public void setLearnerOpts(java.lang.String opts)

Set the parameters to pass to the scikit-learn scheme

Parameters:

opts - the parameters to use

setBatchSize

public void setBatchSize(java.lang.String size)

Specified by:

setBatchSize in interface weka.core.BatchPredictor

Overrides:

setBatchSize in class weka.classifiers.AbstractClassifier

getBatchSize

@OptionMetadata(displayName="Batch size",
                description="The preferred number of instances to transfer into python for prediction\n(if operatingin batch prediction mode). More or fewer instances than this will be accepted.",
                commandLineParamName="batch",
                commandLineParamSynopsis="-batch <batch size>",
                displayOrder=4)
public java.lang.String getBatchSize()

Specified by:

getBatchSize in interface weka.core.BatchPredictor

Overrides:

getBatchSize in class weka.classifiers.AbstractClassifier

implementsMoreEfficientBatchPrediction

public boolean implementsMoreEfficientBatchPrediction()

Returns true, as we send entire test sets over to python for prediction

Specified by:

implementsMoreEfficientBatchPrediction in interface weka.core.BatchPredictor

Overrides:

implementsMoreEfficientBatchPrediction in class weka.classifiers.AbstractClassifier

Returns:

true

setContinueOnSysErr

public void setContinueOnSysErr(boolean c)

Set whether to try and continue after seeing output on the sys error stream. Some schemes write warnings (rather than errors) to sys error.

Parameters:

c - true if we should try to continue after seeing output on the sys error stream

getContinueOnSysErr

@OptionMetadata(displayName="Try to continue after sys err output from script",
                description="Try to continue after sys err output from script.\nSome schemes report warnings to the system error stream.",
                displayOrder=5,
                commandLineParamName="continue-on-err",
                commandLineParamSynopsis="-continue-on-err",
                commandLineParamIsFlag=true)
public boolean getContinueOnSysErr()

Get whether to try and continue after seeing output on the sys error stream. Some schemes write warnings (rather than errors) to sys error.

Returns:

true if we should try to continue after seeing output on the sys error stream

setDontFetchModelFromPython

public void setDontFetchModelFromPython(boolean dontFetchModelFromPython)

If true then don't retrieve the model from python. This can speed up cross-validation but will prevent the trained classifier from being used after deserialization. May also be necessary if a model in python is very large and exceeds the current transfer size (Integer.MAX_VALUE bytes, about 2.1Gb).

Parameters:

dontFetchModelFromPython - true to not fetch the model from python.

getDontFetchModelFromPython

@OptionMetadata(displayName="Don\'t retrieve model from python",
                description="Don\'t retrieve the model from python - speeds up cross-validation,\nbut prevents this classifier from being used after deserialization.\nSome models in python (e.g. large random forests) may exceed the maximum size for transfer\n(currently Integer.MAX_VALUE bytes)",
                displayOrder=6,
                commandLineParamName="dont-fetch-model",
                commandLineParamSynopsis="-dont-fetch-model",
                commandLineParamIsFlag=true)
public boolean getDontFetchModelFromPython()

If true then don't retrieve the model from python. This can speed up cross-validation but will prevent the trained classifier from being used after deserialization. May also be necessary if a model in python is very large and exceeds the current transfer size (Integer.MAX_VALUE bytes, about 2.1Gb).

Returns:

true if not fetching the model from python.

setPythonCommand

public void setPythonCommand(java.lang.String pyCommand)

Set the python command to use. Empty string or "default" indicate that the python present in the PATH should be used.

Parameters:

pyCommand - the path to the python executable (or empty string/"default" to use python in the PATH)

getPythonCommand

@OptionMetadata(displayName="Python command ",
                description="Path to python executable (\'default\' to use python in the PATH)",
                commandLineParamName="py-command",
                commandLineParamSynopsis="-py-command <path to python executable>",
                displayOrder=7)
public java.lang.String getPythonCommand()

Get the python command to use. Empty string or "default" indicate that the python present in the PATH should be used.

Returns:

the path to the python executable (or empty string/"default" to use python in the PATH)

setPythonPath

public void setPythonPath(java.lang.String pythonPath)

Set optional entries to prepend to the PATH so that python can execute correctly. Only applies when not using a default python.

Parameters:

pythonPath - additional entries to prepend to the PATH

getPythonPath

@OptionMetadata(displayName="Python path",
                description="Optional elements to prepend to the PATH so that python can execute correctly (\'default\' to use PATH as-is)",
                commandLineParamName="py-path",
                commandLineParamSynopsis="-py-path <path>",
                displayOrder=8)
public java.lang.String getPythonPath()

Get optional entries to prepend to the PATH so that python can execute correctly. Only applies when not using a default python.

Returns:

additional entries to prepend to the PATH

setServerID

public void setServerID(java.lang.String serverID)

Set an optional server name by which to identify the python server to use. Can be used share a given server amongst selected instances or reserve a server specifically for this classifier. Python command + server name uniquely identifies the server to use.

Parameters:

serverID - the name of the server to use (or none for no specific server name).

getServerID

@OptionMetadata(displayName="Server name/ID",
                description="Optional name to identify this server, can be used to share a given server instance - default = \'none\' (i.e. no server name)",
                commandLineParamName="server-name",
                commandLineParamSynopsis="-server-name <server name | none>",
                displayOrder=9)
public java.lang.String getServerID()

Get an optional server name by which to identify the python server to use. Can be used share a given server amongst selected instances or reserve a server specifically for this classifier. Python command + server name uniquely identifies the server to use.

Returns:

the name of the server to use (or none) for no specific server name.

buildClassifier

public void buildClassifier(weka.core.Instances data)
                     throws java.lang.Exception

Build the classifier

Specified by:

buildClassifier in interface weka.classifiers.Classifier

Parameters:

data - set of instances serving as training data

Throws:

java.lang.Exception - if a problem occurs

distributionForInstance

public double[] distributionForInstance(weka.core.Instance instance)
                                 throws java.lang.Exception

Return the predicted probabilities for the supplied instance

Specified by:

distributionForInstance in interface weka.classifiers.Classifier

Overrides:

distributionForInstance in class weka.classifiers.AbstractClassifier

Parameters:

instance - the instance to be classified

Returns:

the predicted probabilities

Throws:

java.lang.Exception - if a problem occurs

distributionsForInstances

public double[][] distributionsForInstances(weka.core.Instances insts)
                                     throws java.lang.Exception

Return the predicted probabilities for the supplied instances

Specified by:

distributionsForInstances in interface weka.core.BatchPredictor

Overrides:

distributionsForInstances in class weka.classifiers.AbstractClassifier

Parameters:

insts - the instances to get predictions for

Returns:

the predicted probabilities for the supplied instances

Throws:

java.lang.Exception - if a problem occurs

toString

public java.lang.String toString()

Get a textual description of this scheme

Overrides:

toString in class java.lang.Object

Returns:

a textual description of this scheme

main

public static void main(java.lang.String[] argv)

Main method for testing this class

Parameters:

argv - command line args