GBT Regression

It supports both continuous and categorical features.

Input

This takes in a DataFrame and performs Logistic Regression

Output

It generates the GBTRegression and passes it to the next Predict and ModelSave Nodes. The input DataFrame is also passed along to the next nodes.

Type

ml-estimator

Class

fire.nodes.ml.NodeGBTRegression

Fields

Name	Title	Description
featuresCol	Features Column	Features column of type vectorUDT for model fitting
labelCol	Label Column	The label column for model fitting
predictionCol	Prediction Column	The prediction column created during model scoring.
splitRatio	Split Ratio	Split Ratio
impurity	Impurity	The Criterion used for information gain calculation
lossType	Loss Function	The Loss function which GBT tries to minimize
maxBins	Max Bins	The maximum number of bins used for discretizing continuous features.Must be >= 2 and >= number of categories in any categorical feature.
maxDepth	Max Depth	The Maximum depth of a tree
maxIter	Max Iterations	The maximum number of iterations(>=0)(a.k.a numtrees)
minInfoGain	Min Information Gain	The Minimum information gain for a split to be considered at a tree node
minInstancesPerNode	Min Instances Per Node	The Minimum number of instances each child must have after split
subsamplingRate	Subsampling Rate	The fraction of the training data used for learning each decision tree.
seed	Seed	The random seed
stepSize	Step Size	Step size (a.k.a. learning rate), The step size to be used for each iteration of optimization.
cacheNodeIds	Cache Node Ids	The caching nodes IDs. Can speed up training of deeper trees.
checkpointInterval	Checkpoint Interval	The checkpoint interval. E.g. 10 means that the cache will get checkpointed every 10 iterations.Set checkpoint interval (>= 1) or disable checkpoint (-1)
maxMemoryInMB	Max memory	Maximum memory in MB allocated to histogram aggregation.
validationIndicatorCol	Validation Indicator Column	Param for name of the column that indicates whether each row is for training or for validation.
featureSubsetStrategy	Feature Subset Strategy	The number of features to consider for splits at each tree node
minWeightFractionPerNode	Min Weight Fraction per Node	Minimum fraction of the weighted sample count that each child must have after split.
weightCol	Weight Column	Param for weight column name. If this is not set or empty, we treat all instance weights as 1.0.
gridSearch	Grid Search
minInfoGainGrid	Min Info Gain Grid Search	Min Info Gain Grid Search
maxBinsGrid	Max Bins Grid Search	Max Bins for Grid Search
maxDepthGrid	Max Depth Grid Search	Regularization Parameters for Grid Search
maxIterGrid	Max Iterations Grid Search	Max Iterations for Grid Search

Details

Gradient-Boosted Trees (GBTs) are ensembles of decision trees. GBTs iteratively train decision trees in order to minimize a loss function.

The spark.ml implementation supports GBTs for binary classification and for regression, using both continuous and categorical features.

More details are available at Apache Spark ML docs page:

http://spark.apache.org/docs/latest/ml-classification-regression.html#gradient-boosted-tree-regression

Examples

Below example is available at : https://spark.apache.org/docs/latest/ml-classification-regression.html#gradient-boosted-tree-regression

import org.apache.spark.ml.Pipeline

import org.apache.spark.ml.evaluation.RegressionEvaluator

import org.apache.spark.ml.feature.VectorIndexer

import org.apache.spark.ml.regression.{GBTRegressionModel, GBTRegressor}

// Load and parse the data file, converting it to a DataFrame.

val data = spark.read.format(“libsvm”).load(“data/mllib/sample_libsvm_data.txt”)

// Automatically identify categorical features, and index them.

// Set maxCategories so features with > 4 distinct values are treated as continuous.

val featureIndexer = new VectorIndexer()

.setInputCol(“features”)

.setOutputCol(“indexedFeatures”)

.setMaxCategories(4)

.fit(data)

// Split the data into training and test sets (30% held out for testing).

val Array(trainingData, testData) = data.randomSplit(Array(0.7, 0.3))

// Train a GBT model.

val gbt = new GBTRegressor()

.setLabelCol(“label”)

.setFeaturesCol(“indexedFeatures”)

.setMaxIter(10)

// Chain indexer and GBT in a Pipeline.

val pipeline = new Pipeline()

.setStages(Array(featureIndexer, gbt))

// Train model. This also runs the indexer.

val model = pipeline.fit(trainingData)

// Make predictions.

val predictions = model.transform(testData)

// Select example rows to display.

predictions.select(“prediction”, “label”, “features”).show(5)

// Select (prediction, true label) and compute test error.

val evaluator = new RegressionEvaluator()

.setLabelCol(“label”)

.setPredictionCol(“prediction”)

.setMetricName(“rmse”)

val rmse = evaluator.evaluate(predictions)

println(s”Root Mean Squared Error (RMSE) on test data = $rmse”)

val gbtModel = model.stages(1).asInstanceOf[GBTRegressionModel]

println(s”Learned regression GBT model:\n ${gbtModel.toDebugString}”)