Class GBTClassifier
- All Implemented Interfaces:
Serializable
,org.apache.spark.internal.Logging
,ClassifierParams
,ProbabilisticClassifierParams
,Params
,HasCheckpointInterval
,HasFeaturesCol
,HasLabelCol
,HasMaxIter
,HasPredictionCol
,HasProbabilityCol
,HasRawPredictionCol
,HasSeed
,HasStepSize
,HasThresholds
,HasValidationIndicatorCol
,HasWeightCol
,PredictorParams
,DecisionTreeParams
,GBTClassifierParams
,GBTParams
,HasVarianceImpurity
,TreeEnsembleClassifierParams
,TreeEnsembleParams
,DefaultParamsWritable
,Identifiable
,MLWritable
The implementation is based upon: J.H. Friedman. "Stochastic Gradient Boosting." 1999.
Notes on Gradient Boosting vs. TreeBoost: - This implementation is for Stochastic Gradient Boosting, not for TreeBoost. - Both algorithms learn tree ensembles by minimizing loss functions. - TreeBoost (Friedman, 1999) additionally modifies the outputs at tree leaf nodes based on the loss function, whereas the original gradient boosting method does not. - We expect to implement TreeBoost in the future: [https://issues.apache.org/jira/browse/SPARK-4240]
- See Also:
- Note:
- Multiclass labels are not currently supported.
-
Nested Class Summary
Nested classes/interfaces inherited from interface org.apache.spark.internal.Logging
org.apache.spark.internal.Logging.LogStringContext, org.apache.spark.internal.Logging.SparkShellLoggingFilter
-
Constructor Summary
-
Method Summary
Modifier and TypeMethodDescriptionfinal BooleanParam
If false, the algorithm will pass trees to executors to match instances with nodes.final IntParam
Param for set checkpoint interval (>= 1) or disable checkpoint (-1).Creates a copy of this instance with the same UID and some extra params.The number of features to consider for splits at each tree node.impurity()
Criterion used for information gain calculation (case-insensitive).leafCol()
Leaf indices column name.static GBTClassifier
lossType()
Loss function which GBT tries to minimize.final IntParam
maxBins()
Maximum number of bins used for discretizing continuous features and for choosing how to split on features at each node.final IntParam
maxDepth()
Maximum depth of the tree (nonnegative).final IntParam
maxIter()
Param for maximum number of iterations (>= 0).final IntParam
Maximum memory in MB allocated to histogram aggregation.final DoubleParam
Minimum information gain for a split to be considered at a tree node.final IntParam
Minimum number of instances each child must have after split.final DoubleParam
Minimum fraction of the weighted sample count that each child must have after split.static MLReader<T>
read()
final LongParam
seed()
Param for random seed.setCacheNodeIds
(boolean value) setCheckpointInterval
(int value) Specifies how often to checkpoint the cached node IDs.setFeatureSubsetStrategy
(String value) setImpurity
(String value) The impurity setting is ignored for GBT models.setLossType
(String value) setMaxBins
(int value) setMaxDepth
(int value) setMaxIter
(int value) setMaxMemoryInMB
(int value) setMinInfoGain
(double value) setMinInstancesPerNode
(int value) setMinWeightFractionPerNode
(double value) setSeed
(long value) setStepSize
(double value) setSubsamplingRate
(double value) setValidationIndicatorCol
(String value) setWeightCol
(String value) Sets the value of paramweightCol()
.final DoubleParam
stepSize()
Param for Step size (a.k.a.final DoubleParam
Fraction of the training data used for learning each decision tree, in range (0, 1].static final String[]
Accessor for supported loss settings: logisticuid()
An immutable unique ID for the object and its derivatives.Param for name of the column that indicates whether each row is for training or for validation.final DoubleParam
Threshold for stopping early when fit with validation is used.Param for weight column name.Methods inherited from class org.apache.spark.ml.classification.ProbabilisticClassifier
probabilityCol, setProbabilityCol, setThresholds, thresholds
Methods inherited from class org.apache.spark.ml.classification.Classifier
rawPredictionCol, setRawPredictionCol
Methods inherited from class org.apache.spark.ml.Predictor
featuresCol, fit, labelCol, predictionCol, setFeaturesCol, setLabelCol, setPredictionCol, transformSchema
Methods inherited from class org.apache.spark.ml.PipelineStage
params
Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
Methods inherited from interface org.apache.spark.ml.tree.DecisionTreeParams
getCacheNodeIds, getLeafCol, getMaxBins, getMaxDepth, getMaxMemoryInMB, getMinInfoGain, getMinInstancesPerNode, getMinWeightFractionPerNode, getOldStrategy, setLeafCol
Methods inherited from interface org.apache.spark.ml.util.DefaultParamsWritable
write
Methods inherited from interface org.apache.spark.ml.tree.GBTClassifierParams
getLossType, getOldLossType
Methods inherited from interface org.apache.spark.ml.tree.GBTParams
getOldBoostingStrategy, getValidationTol
Methods inherited from interface org.apache.spark.ml.param.shared.HasCheckpointInterval
getCheckpointInterval
Methods inherited from interface org.apache.spark.ml.param.shared.HasFeaturesCol
featuresCol, getFeaturesCol
Methods inherited from interface org.apache.spark.ml.param.shared.HasLabelCol
getLabelCol, labelCol
Methods inherited from interface org.apache.spark.ml.param.shared.HasMaxIter
getMaxIter
Methods inherited from interface org.apache.spark.ml.param.shared.HasPredictionCol
getPredictionCol, predictionCol
Methods inherited from interface org.apache.spark.ml.param.shared.HasProbabilityCol
getProbabilityCol, probabilityCol
Methods inherited from interface org.apache.spark.ml.param.shared.HasRawPredictionCol
getRawPredictionCol, rawPredictionCol
Methods inherited from interface org.apache.spark.ml.param.shared.HasStepSize
getStepSize
Methods inherited from interface org.apache.spark.ml.param.shared.HasThresholds
getThresholds, thresholds
Methods inherited from interface org.apache.spark.ml.param.shared.HasValidationIndicatorCol
getValidationIndicatorCol
Methods inherited from interface org.apache.spark.ml.tree.HasVarianceImpurity
getImpurity, getOldImpurity
Methods inherited from interface org.apache.spark.ml.param.shared.HasWeightCol
getWeightCol
Methods inherited from interface org.apache.spark.ml.util.Identifiable
toString
Methods inherited from interface org.apache.spark.internal.Logging
initializeForcefully, initializeLogIfNecessary, initializeLogIfNecessary, initializeLogIfNecessary$default$2, isTraceEnabled, log, logDebug, logDebug, logDebug, logDebug, logError, logError, logError, logError, logInfo, logInfo, logInfo, logInfo, logName, LogStringContext, logTrace, logTrace, logTrace, logTrace, logWarning, logWarning, logWarning, logWarning, org$apache$spark$internal$Logging$$log_, org$apache$spark$internal$Logging$$log__$eq, withLogContext
Methods inherited from interface org.apache.spark.ml.util.MLWritable
save
Methods inherited from interface org.apache.spark.ml.param.Params
clear, copyValues, defaultCopy, defaultParamMap, explainParam, explainParams, extractParamMap, extractParamMap, get, getDefault, getOrDefault, getParam, hasDefault, hasParam, isDefined, isSet, onParamChange, paramMap, params, set, set, set, setDefault, setDefault, shouldOwn
Methods inherited from interface org.apache.spark.ml.tree.TreeEnsembleClassifierParams
validateAndTransformSchema
Methods inherited from interface org.apache.spark.ml.tree.TreeEnsembleParams
getFeatureSubsetStrategy, getOldStrategy, getSubsamplingRate
-
Constructor Details
-
GBTClassifier
-
GBTClassifier
public GBTClassifier()
-
-
Method Details
-
supportedLossTypes
Accessor for supported loss settings: logistic -
load
-
read
-
lossType
Description copied from interface:GBTClassifierParams
Loss function which GBT tries to minimize. (case-insensitive) Supported: "logistic" (default = logistic)- Specified by:
lossType
in interfaceGBTClassifierParams
- Returns:
- (undocumented)
-
impurity
Description copied from interface:HasVarianceImpurity
Criterion used for information gain calculation (case-insensitive). This impurity type is used in DecisionTreeRegressor, RandomForestRegressor, GBTRegressor and GBTClassifier (since GBTClassificationModel is internally composed of DecisionTreeRegressionModels). Supported: "variance". (default = variance)- Specified by:
impurity
in interfaceHasVarianceImpurity
- Returns:
- (undocumented)
-
validationTol
Description copied from interface:GBTParams
Threshold for stopping early when fit with validation is used. (This parameter is ignored when fit without validation is used.) The decision to stop early is decided based on this logic: If the current loss on the validation set is greater than 0.01, the diff of validation error is compared to relative tolerance which is validationTol * (current loss on the validation set). If the current loss on the validation set is less than or equal to 0.01, the diff of validation error is compared to absolute tolerance which is validationTol * 0.01.- Specified by:
validationTol
in interfaceGBTParams
- Returns:
- (undocumented)
- See Also:
-
stepSize
Description copied from interface:GBTParams
Param for Step size (a.k.a. learning rate) in interval (0, 1] for shrinking the contribution of each estimator. (default = 0.1)- Specified by:
stepSize
in interfaceGBTParams
- Specified by:
stepSize
in interfaceHasStepSize
- Returns:
- (undocumented)
-
validationIndicatorCol
Description copied from interface:HasValidationIndicatorCol
Param for name of the column that indicates whether each row is for training or for validation. False indicates training; true indicates validation..- Specified by:
validationIndicatorCol
in interfaceHasValidationIndicatorCol
- Returns:
- (undocumented)
-
maxIter
Description copied from interface:HasMaxIter
Param for maximum number of iterations (>= 0).- Specified by:
maxIter
in interfaceHasMaxIter
- Returns:
- (undocumented)
-
subsamplingRate
Description copied from interface:TreeEnsembleParams
Fraction of the training data used for learning each decision tree, in range (0, 1]. (default = 1.0)- Specified by:
subsamplingRate
in interfaceTreeEnsembleParams
- Returns:
- (undocumented)
-
featureSubsetStrategy
Description copied from interface:TreeEnsembleParams
The number of features to consider for splits at each tree node. Supported options: - "auto": Choose automatically for task: If numTrees == 1, set to "all." If numTrees greater than 1 (forest), set to "sqrt" for classification and to "onethird" for regression. - "all": use all features - "onethird": use 1/3 of the features - "sqrt": use sqrt(number of features) - "log2": use log2(number of features) - "n": when n is in the range (0, 1.0], use n * number of features. When n is in the range (1, number of features), use n features. (default = "auto")These various settings are based on the following references: - log2: tested in Breiman (2001) - sqrt: recommended by Breiman manual for random forests - The defaults of sqrt (classification) and onethird (regression) match the R randomForest package.
- Specified by:
featureSubsetStrategy
in interfaceTreeEnsembleParams
- Returns:
- (undocumented)
- See Also:
-
leafCol
Description copied from interface:DecisionTreeParams
Leaf indices column name. Predicted leaf index of each instance in each tree by preorder. (default = "")- Specified by:
leafCol
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
maxDepth
Description copied from interface:DecisionTreeParams
Maximum depth of the tree (nonnegative). E.g., depth 0 means 1 leaf node; depth 1 means 1 internal node + 2 leaf nodes. (default = 5)- Specified by:
maxDepth
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
maxBins
Description copied from interface:DecisionTreeParams
Maximum number of bins used for discretizing continuous features and for choosing how to split on features at each node. More bins give higher granularity. Must be at least 2 and at least number of categories in any categorical feature. (default = 32)- Specified by:
maxBins
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
minInstancesPerNode
Description copied from interface:DecisionTreeParams
Minimum number of instances each child must have after split. If a split causes the left or right child to have fewer than minInstancesPerNode, the split will be discarded as invalid. Must be at least 1. (default = 1)- Specified by:
minInstancesPerNode
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
minWeightFractionPerNode
Description copied from interface:DecisionTreeParams
Minimum fraction of the weighted sample count that each child must have after split. If a split causes the fraction of the total weight in the left or right child to be less than minWeightFractionPerNode, the split will be discarded as invalid. Should be in the interval [0.0, 0.5). (default = 0.0)- Specified by:
minWeightFractionPerNode
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
minInfoGain
Description copied from interface:DecisionTreeParams
Minimum information gain for a split to be considered at a tree node. Should be at least 0.0. (default = 0.0)- Specified by:
minInfoGain
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
maxMemoryInMB
Description copied from interface:DecisionTreeParams
Maximum memory in MB allocated to histogram aggregation. If too small, then 1 node will be split per iteration, and its aggregates may exceed this size. (default = 256 MB)- Specified by:
maxMemoryInMB
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
cacheNodeIds
Description copied from interface:DecisionTreeParams
If false, the algorithm will pass trees to executors to match instances with nodes. If true, the algorithm will cache node IDs for each instance. Caching can speed up training of deeper trees. Users can set how often should the cache be checkpointed or disable it by setting checkpointInterval. (default = false)- Specified by:
cacheNodeIds
in interfaceDecisionTreeParams
- Returns:
- (undocumented)
-
weightCol
Description copied from interface:HasWeightCol
Param for weight column name. If this is not set or empty, we treat all instance weights as 1.0.- Specified by:
weightCol
in interfaceHasWeightCol
- Returns:
- (undocumented)
-
seed
Description copied from interface:HasSeed
Param for random seed. -
checkpointInterval
Description copied from interface:HasCheckpointInterval
Param for set checkpoint interval (>= 1) or disable checkpoint (-1). E.g. 10 means that the cache will get checkpointed every 10 iterations. Note: this setting will be ignored if the checkpoint directory is not set in the SparkContext.- Specified by:
checkpointInterval
in interfaceHasCheckpointInterval
- Returns:
- (undocumented)
-
uid
Description copied from interface:Identifiable
An immutable unique ID for the object and its derivatives.- Specified by:
uid
in interfaceIdentifiable
- Returns:
- (undocumented)
-
setMaxDepth
-
setMaxBins
-
setMinInstancesPerNode
-
setMinWeightFractionPerNode
-
setMinInfoGain
-
setMaxMemoryInMB
-
setCacheNodeIds
-
setCheckpointInterval
Specifies how often to checkpoint the cached node IDs. E.g. 10 means that the cache will get checkpointed every 10 iterations. This is only used if cacheNodeIds is true and if the checkpoint directory is set inSparkContext
. Must be at least 1. (default = 10)- Parameters:
value
- (undocumented)- Returns:
- (undocumented)
-
setImpurity
The impurity setting is ignored for GBT models. Individual trees are built using impurity "Variance."- Parameters:
value
- (undocumented)- Returns:
- (undocumented)
-
setSubsamplingRate
-
setSeed
-
setMaxIter
-
setStepSize
-
setFeatureSubsetStrategy
-
setLossType
-
setValidationIndicatorCol
-
setWeightCol
Sets the value of paramweightCol()
. If this is not set or empty, we treat all instance weights as 1.0. By default the weightCol is not set, so all instances have weight 1.0.- Parameters:
value
- (undocumented)- Returns:
- (undocumented)
-
copy
Description copied from interface:Params
Creates a copy of this instance with the same UID and some extra params. Subclasses should implement this method and set the return type properly. SeedefaultCopy()
.- Specified by:
copy
in interfaceParams
- Specified by:
copy
in classPredictor<Vector,
GBTClassifier, GBTClassificationModel> - Parameters:
extra
- (undocumented)- Returns:
- (undocumented)
-