Package org.apache.spark.ml.regression

Class GBTRegressor

Object
org.apache.spark.ml.PipelineStage
org.apache.spark.ml.Estimator<M>
org.apache.spark.ml.Predictor<FeaturesType,Learner,M>
org.apache.spark.ml.regression.Regressor<Vector,GBTRegressor,GBTRegressionModel>
org.apache.spark.ml.regression.GBTRegressor
All Implemented Interfaces:
Serializable, org.apache.spark.internal.Logging, Params, HasCheckpointInterval, HasFeaturesCol, HasLabelCol, HasMaxIter, HasPredictionCol, HasSeed, HasStepSize, HasValidationIndicatorCol, HasWeightCol, PredictorParams, DecisionTreeParams, GBTParams, GBTRegressorParams, HasVarianceImpurity, TreeEnsembleParams, TreeEnsembleRegressorParams, TreeRegressorParams, DefaultParamsWritable, Identifiable, MLWritable
public class GBTRegressor extends Regressor<Vector,GBTRegressor,GBTRegressionModel> implements GBTRegressorParams, DefaultParamsWritable, org.apache.spark.internal.Logging
Gradient-Boosted Trees (GBTs) learning algorithm for regression. It supports both continuous and categorical features.

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. - When the loss is SquaredError, these methods give the same result, but they could differ for other loss functions. - We expect to implement TreeBoost in the future: [https://issues.apache.org/jira/browse/SPARK-4240]

See Also:

  • Constructor Details

    • GBTRegressor

      public GBTRegressor(String uid)

    • GBTRegressor

      public GBTRegressor()

  • Method Details

    • supportedLossTypes

      public static final String[] supportedLossTypes()
      Accessor for supported loss settings: squared (L2), absolute (L1)

    • load

      public static GBTRegressor load(String path)

    • read

      public static MLReader<T> read()

    • lossType

      public Param<String> lossType()
      Description copied from interface: GBTRegressorParams
      Loss function which GBT tries to minimize. (case-insensitive) Supported: "squared" (L2) and "absolute" (L1) (default = squared)
      Specified by:
      lossType in interface GBTRegressorParams
      Returns:
      (undocumented)

    • impurity

      public final Param<String> 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 interface HasVarianceImpurity
      Returns:
      (undocumented)

    • validationTol

      public final DoubleParam 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 interface GBTParams
      Returns:
      (undocumented)
      See Also:

    • stepSize

      public final DoubleParam 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 interface GBTParams
      Specified by:
      stepSize in interface HasStepSize
      Returns:
      (undocumented)

    • validationIndicatorCol

      public final Param<String> 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 interface HasValidationIndicatorCol
      Returns:
      (undocumented)

    • maxIter

      public final IntParam maxIter()
      Description copied from interface: HasMaxIter
      Param for maximum number of iterations (&gt;= 0).
      Specified by:
      maxIter in interface HasMaxIter
      Returns:
      (undocumented)

    • subsamplingRate

      public final DoubleParam 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 interface TreeEnsembleParams
      Returns:
      (undocumented)

    • featureSubsetStrategy

      public final Param<String> 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 interface TreeEnsembleParams
      Returns:
      (undocumented)
      See Also:

    • leafCol

      public final Param<String> 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • maxDepth

      public final IntParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • maxBins

      public final IntParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • minInstancesPerNode

      public final IntParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • minWeightFractionPerNode

      public final DoubleParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • minInfoGain

      public final DoubleParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • maxMemoryInMB

      public final IntParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • cacheNodeIds

      public final BooleanParam 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 interface DecisionTreeParams
      Returns:
      (undocumented)

    • weightCol

      public final Param<String> 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 interface HasWeightCol
      Returns:
      (undocumented)

    • seed

      public final LongParam seed()
      Description copied from interface: HasSeed
      Param for random seed.
      Specified by:
      seed in interface HasSeed
      Returns:
      (undocumented)

    • checkpointInterval

      public final IntParam checkpointInterval()
      Description copied from interface: HasCheckpointInterval
      Param for set checkpoint interval (&gt;= 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 interface HasCheckpointInterval
      Returns:
      (undocumented)

    • uid

      public String uid()
      Description copied from interface: Identifiable
      An immutable unique ID for the object and its derivatives.
      Specified by:
      uid in interface Identifiable
      Returns:
      (undocumented)

    • setMaxDepth

      public GBTRegressor setMaxDepth(int value)

    • setMaxBins

      public GBTRegressor setMaxBins(int value)

    • setMinInstancesPerNode

      public GBTRegressor setMinInstancesPerNode(int value)

    • setMinWeightFractionPerNode

      public GBTRegressor setMinWeightFractionPerNode(double value)

    • setMinInfoGain

      public GBTRegressor setMinInfoGain(double value)

    • setMaxMemoryInMB

      public GBTRegressor setMaxMemoryInMB(int value)

    • setCacheNodeIds

      public GBTRegressor setCacheNodeIds(boolean value)

    • setCheckpointInterval

      public GBTRegressor setCheckpointInterval(int value)
      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 in SparkContext. Must be at least 1. (default = 10)
      Parameters:
      value - (undocumented)
      Returns:
      (undocumented)

    • setImpurity

      public GBTRegressor setImpurity(String value)
      The impurity setting is ignored for GBT models. Individual trees are built using impurity "Variance."

      Parameters:
      value - (undocumented)
      Returns:
      (undocumented)

    • setSubsamplingRate

      public GBTRegressor setSubsamplingRate(double value)

    • setSeed

      public GBTRegressor setSeed(long value)

    • setMaxIter

      public GBTRegressor setMaxIter(int value)

    • setStepSize

      public GBTRegressor setStepSize(double value)

    • setLossType

      public GBTRegressor setLossType(String value)

    • setFeatureSubsetStrategy

      public GBTRegressor setFeatureSubsetStrategy(String value)

    • setValidationIndicatorCol

      public GBTRegressor setValidationIndicatorCol(String value)

    • setWeightCol

      public GBTRegressor setWeightCol(String value)
      Sets the value of param weightCol(). 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

      public GBTRegressor copy(ParamMap extra)
      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. See defaultCopy().
      Specified by:
      copy in interface Params
      Specified by:
      copy in class Predictor<Vector,GBTRegressor,GBTRegressionModel>
      Parameters:
      extra - (undocumented)
      Returns:
      (undocumented)