diff --git a/R/pkg/inst/tests/testthat/test_mllib.R b/R/pkg/inst/tests/testthat/test_mllib.R index 33e9d0d267ac5a8036a2994e517acac967b2eb46..b76f75dbdc682eb149e3efed4424fe173597ea53 100644 --- a/R/pkg/inst/tests/testthat/test_mllib.R +++ b/R/pkg/inst/tests/testthat/test_mllib.R @@ -935,6 +935,10 @@ test_that("spark.randomForest Classification", { expect_equal(stats$numTrees, 20) expect_error(capture.output(stats), NA) expect_true(length(capture.output(stats)) > 6) + # Test string prediction values + predictions <- collect(predict(model, data))$prediction + expect_equal(length(grep("setosa", predictions)), 50) + expect_equal(length(grep("versicolor", predictions)), 50) modelPath <- tempfile(pattern = "spark-randomForestClassification", fileext = ".tmp") write.ml(model, modelPath) @@ -947,6 +951,26 @@ test_that("spark.randomForest Classification", { expect_equal(stats$numClasses, stats2$numClasses) unlink(modelPath) + + # Test numeric response variable + labelToIndex <- function(species) { + switch(as.character(species), + setosa = 0.0, + versicolor = 1.0, + virginica = 2.0 + ) + } + iris$NumericSpecies <- lapply(iris$Species, labelToIndex) + data <- suppressWarnings(createDataFrame(iris[-5])) + model <- spark.randomForest(data, NumericSpecies ~ Petal_Length + Petal_Width, "classification", + maxDepth = 5, maxBins = 16) + stats <- summary(model) + expect_equal(stats$numFeatures, 2) + expect_equal(stats$numTrees, 20) + # Test numeric prediction values + predictions <- collect(predict(model, data))$prediction + expect_equal(length(grep("1.0", predictions)), 50) + expect_equal(length(grep("2.0", predictions)), 50) }) test_that("spark.gbt", { diff --git a/mllib/src/main/scala/org/apache/spark/ml/r/RandomForestClassificationWrapper.scala b/mllib/src/main/scala/org/apache/spark/ml/r/RandomForestClassificationWrapper.scala index 6947ba7e7597ab0384a2a7cc0ffc350a3fe99e7d..31f846dc6cfec0ba34909f2502d42e8231400e60 100644 --- a/mllib/src/main/scala/org/apache/spark/ml/r/RandomForestClassificationWrapper.scala +++ b/mllib/src/main/scala/org/apache/spark/ml/r/RandomForestClassificationWrapper.scala @@ -23,9 +23,9 @@ import org.json4s.JsonDSL._ import org.json4s.jackson.JsonMethods._ import org.apache.spark.ml.{Pipeline, PipelineModel} -import org.apache.spark.ml.attribute.AttributeGroup +import org.apache.spark.ml.attribute.{Attribute, AttributeGroup, NominalAttribute} import org.apache.spark.ml.classification.{RandomForestClassificationModel, RandomForestClassifier} -import org.apache.spark.ml.feature.RFormula +import org.apache.spark.ml.feature.{IndexToString, RFormula} import org.apache.spark.ml.linalg.Vector import org.apache.spark.ml.util._ import org.apache.spark.sql.{DataFrame, Dataset} @@ -35,6 +35,8 @@ private[r] class RandomForestClassifierWrapper private ( val formula: String, val features: Array[String]) extends MLWritable { + import RandomForestClassifierWrapper._ + private val rfcModel: RandomForestClassificationModel = pipeline.stages(1).asInstanceOf[RandomForestClassificationModel] @@ -46,7 +48,9 @@ private[r] class RandomForestClassifierWrapper private ( def summary: String = rfcModel.toDebugString def transform(dataset: Dataset[_]): DataFrame = { - pipeline.transform(dataset).drop(rfcModel.getFeaturesCol) + pipeline.transform(dataset) + .drop(PREDICTED_LABEL_INDEX_COL) + .drop(rfcModel.getFeaturesCol) } override def write: MLWriter = new @@ -54,6 +58,10 @@ private[r] class RandomForestClassifierWrapper private ( } private[r] object RandomForestClassifierWrapper extends MLReadable[RandomForestClassifierWrapper] { + + val PREDICTED_LABEL_INDEX_COL = "pred_label_idx" + val PREDICTED_LABEL_COL = "prediction" + def fit( // scalastyle:ignore data: DataFrame, formula: String, @@ -73,6 +81,7 @@ private[r] object RandomForestClassifierWrapper extends MLReadable[RandomForestC val rFormula = new RFormula() .setFormula(formula) + .setForceIndexLabel(true) RWrapperUtils.checkDataColumns(rFormula, data) val rFormulaModel = rFormula.fit(data) @@ -82,6 +91,11 @@ private[r] object RandomForestClassifierWrapper extends MLReadable[RandomForestC .attributes.get val features = featureAttrs.map(_.name.get) + // get label names from output schema + val labelAttr = Attribute.fromStructField(schema(rFormulaModel.getLabelCol)) + .asInstanceOf[NominalAttribute] + val labels = labelAttr.values.get + // assemble and fit the pipeline val rfc = new RandomForestClassifier() .setMaxDepth(maxDepth) @@ -97,10 +111,16 @@ private[r] object RandomForestClassifierWrapper extends MLReadable[RandomForestC .setCacheNodeIds(cacheNodeIds) .setProbabilityCol(probabilityCol) .setFeaturesCol(rFormula.getFeaturesCol) + .setPredictionCol(PREDICTED_LABEL_INDEX_COL) if (seed != null && seed.length > 0) rfc.setSeed(seed.toLong) + val idxToStr = new IndexToString() + .setInputCol(PREDICTED_LABEL_INDEX_COL) + .setOutputCol(PREDICTED_LABEL_COL) + .setLabels(labels) + val pipeline = new Pipeline() - .setStages(Array(rFormulaModel, rfc)) + .setStages(Array(rFormulaModel, rfc, idxToStr)) .fit(data) new RandomForestClassifierWrapper(pipeline, formula, features)