org.apache.spark.mllib.tree.model

Class GradientBoostedTreesModel

java.lang.Object

org.apache.spark.mllib.tree.model.GradientBoostedTreesModel

All Implemented Interfaces:

java.io.Serializable, Saveable

public class GradientBoostedTreesModel
extends java.lang.Object
implements Saveable

Represents a gradient boosted trees model.

param: algo algorithm for the ensemble model, either Classification or Regression param: trees tree ensembles param: treeWeights tree ensemble weights

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
GradientBoostedTreesModel(scala.Enumeration.Value algo, DecisionTreeModel[] trees, double[] treeWeights)

Method Summary

Methods

Modifier and Type	Method and Description
scala.Enumeration.Value	algo()
protected scala.Enumeration.Value	combiningStrategy()
static RDD<scala.Tuple2<java.lang.Object,java.lang.Object>>	computeInitialPredictionAndError(RDD<LabeledPoint> data, double initTreeWeight, DecisionTreeModel initTree, Loss loss) :: DeveloperApi :: Compute the initial predictions and errors for a dataset for the first iteration of gradient boosting.
double[]	evaluateEachIteration(RDD<LabeledPoint> data, Loss loss) Method to compute error or loss for every iteration of gradient boosting.
protected java.lang.String	formatVersion() Current version of model save/load format.
static GradientBoostedTreesModel	load(SparkContext sc, java.lang.String path)
int	numTrees() Get number of trees in ensemble.
JavaRDD<java.lang.Double>	predict(JavaRDD<Vector> features) Java-friendly version of TreeEnsembleModel.predict(org.apache.spark.mllib.linalg.Vector).
RDD<java.lang.Object>	predict(RDD<Vector> features) Predict values for the given data set.
double	predict(Vector features) Predict values for a single data point using the model trained.
void	save(SparkContext sc, java.lang.String path) Save this model to the given path.
java.lang.String	toDebugString() Print the full model to a string.
java.lang.String	toString() Print a summary of the model.
int	totalNumNodes() Get total number of nodes, summed over all trees in the ensemble.
DecisionTreeModel[]	trees()
double[]	treeWeights()
static RDD<scala.Tuple2<java.lang.Object,java.lang.Object>>	updatePredictionError(RDD<LabeledPoint> data, RDD<scala.Tuple2<java.lang.Object,java.lang.Object>> predictionAndError, double treeWeight, DecisionTreeModel tree, Loss loss) :: DeveloperApi :: Update a zipped predictionError RDD (as obtained with computeInitialPredictionAndError)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Detail

GradientBoostedTreesModel

public GradientBoostedTreesModel(scala.Enumeration.Value algo,
                         DecisionTreeModel[] trees,
                         double[] treeWeights)

Method Detail

computeInitialPredictionAndError

public static RDD<scala.Tuple2<java.lang.Object,java.lang.Object>> computeInitialPredictionAndError(RDD<LabeledPoint> data,
                                                                                    double initTreeWeight,
                                                                                    DecisionTreeModel initTree,
                                                                                    Loss loss)

:: DeveloperApi :: Compute the initial predictions and errors for a dataset for the first iteration of gradient boosting.

Parameters:

data: - training data.

initTreeWeight: - learning rate assigned to the first tree.

initTree: - first DecisionTreeModel.

loss: - evaluation metric.

Returns:

a RDD with each element being a zip of the prediction and error corresponding to every sample.

updatePredictionError

public static RDD<scala.Tuple2<java.lang.Object,java.lang.Object>> updatePredictionError(RDD<LabeledPoint> data,
                                                                         RDD<scala.Tuple2<java.lang.Object,java.lang.Object>> predictionAndError,
                                                                         double treeWeight,
                                                                         DecisionTreeModel tree,
                                                                         Loss loss)

:: DeveloperApi :: Update a zipped predictionError RDD (as obtained with computeInitialPredictionAndError)

Parameters:

data: - training data.

predictionAndError: - predictionError RDD

treeWeight: - Learning rate.

tree: - Tree using which the prediction and error should be updated.

loss: - evaluation metric.

Returns:

a RDD with each element being a zip of the prediction and error corresponding to each sample.

load

public static GradientBoostedTreesModel load(SparkContext sc,
                             java.lang.String path)

Parameters:

sc - Spark context used for loading model files.

path - Path specifying the directory to which the model was saved.

Returns:

Model instance

algo

public scala.Enumeration.Value algo()

trees

public DecisionTreeModel[] trees()

treeWeights

public double[] treeWeights()

save

public void save(SparkContext sc,
        java.lang.String path)

Description copied from interface: Saveable

Save this model to the given path.

This saves: - human-readable (JSON) model metadata to path/metadata/ - Parquet formatted data to path/data/

The model may be loaded using Loader.load.

Specified by:

save in interface Saveable

Parameters:

sc - Spark context used to save model data.

path - Path specifying the directory in which to save this model. If the directory already exists, this method throws an exception.

evaluateEachIteration

public double[] evaluateEachIteration(RDD<LabeledPoint> data,
                             Loss loss)

Method to compute error or loss for every iteration of gradient boosting.

Parameters:

data - RDD of LabeledPoint

loss - evaluation metric.

Returns:

an array with index i having the losses or errors for the ensemble containing the first i+1 trees

formatVersion

protected java.lang.String formatVersion()

Description copied from interface: Saveable

Current version of model save/load format.

Specified by:

formatVersion in interface Saveable

combiningStrategy

protected scala.Enumeration.Value combiningStrategy()

predict

public double predict(Vector features)

Predict values for a single data point using the model trained.

Parameters:

features - array representing a single data point

Returns:

predicted category from the trained model

predict

public RDD<java.lang.Object> predict(RDD<Vector> features)

Predict values for the given data set.

Parameters:

features - RDD representing data points to be predicted

Returns:

RDD[Double] where each entry contains the corresponding prediction

predict

public JavaRDD<java.lang.Double> predict(JavaRDD<Vector> features)

Java-friendly version of TreeEnsembleModel.predict(org.apache.spark.mllib.linalg.Vector).

Parameters:

features - (undocumented)

Returns:

(undocumented)

toString

public java.lang.String toString()

Print a summary of the model.

Overrides:

toString in class java.lang.Object

Returns:

(undocumented)

toDebugString

public java.lang.String toDebugString()

Print the full model to a string.

Returns:

(undocumented)

numTrees

public int numTrees()

Get number of trees in ensemble.

Returns:

(undocumented)

totalNumNodes

public int totalNumNodes()

Get total number of nodes, summed over all trees in the ensemble.

Returns:

(undocumented)