diff options
Diffstat (limited to 'mllib')
3 files changed, 82 insertions, 2 deletions
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala index 7d912737b8..1d5d3762ed 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/api/python/PythonMLLibAPI.scala @@ -19,6 +19,8 @@ package org.apache.spark.mllib.api.python import java.nio.{ByteBuffer, ByteOrder} +import scala.collection.JavaConverters._ + import org.apache.spark.annotation.DeveloperApi import org.apache.spark.api.java.{JavaRDD, JavaSparkContext} import org.apache.spark.mllib.classification._ @@ -29,6 +31,11 @@ import org.apache.spark.mllib.linalg.{Matrix, SparseVector, Vector, Vectors} import org.apache.spark.mllib.random.{RandomRDDGenerators => RG} import org.apache.spark.mllib.recommendation._ import org.apache.spark.mllib.regression._ +import org.apache.spark.mllib.tree.configuration.Algo._ +import org.apache.spark.mllib.tree.configuration.Strategy +import org.apache.spark.mllib.tree.DecisionTree +import org.apache.spark.mllib.tree.impurity.{Entropy, Gini, Impurity, Variance} +import org.apache.spark.mllib.tree.model.DecisionTreeModel import org.apache.spark.mllib.stat.Statistics import org.apache.spark.mllib.stat.correlation.CorrelationNames import org.apache.spark.mllib.util.MLUtils @@ -473,6 +480,76 @@ class PythonMLLibAPI extends Serializable { } /** + * Java stub for Python mllib DecisionTree.train(). + * This stub returns a handle to the Java object instead of the content of the Java object. + * Extra care needs to be taken in the Python code to ensure it gets freed on exit; + * see the Py4J documentation. + * @param dataBytesJRDD Training data + * @param categoricalFeaturesInfoJMap Categorical features info, as Java map + */ + def trainDecisionTreeModel( + dataBytesJRDD: JavaRDD[Array[Byte]], + algoStr: String, + numClasses: Int, + categoricalFeaturesInfoJMap: java.util.Map[Int, Int], + impurityStr: String, + maxDepth: Int, + maxBins: Int): DecisionTreeModel = { + + val data = dataBytesJRDD.rdd.map(deserializeLabeledPoint) + + val algo: Algo = algoStr match { + case "classification" => Classification + case "regression" => Regression + case _ => throw new IllegalArgumentException(s"Bad algoStr parameter: $algoStr") + } + val impurity: Impurity = impurityStr match { + case "gini" => Gini + case "entropy" => Entropy + case "variance" => Variance + case _ => throw new IllegalArgumentException(s"Bad impurityStr parameter: $impurityStr") + } + + val strategy = new Strategy( + algo = algo, + impurity = impurity, + maxDepth = maxDepth, + numClassesForClassification = numClasses, + maxBins = maxBins, + categoricalFeaturesInfo = categoricalFeaturesInfoJMap.asScala.toMap) + + DecisionTree.train(data, strategy) + } + + /** + * Predict the label of the given data point. + * This is a Java stub for python DecisionTreeModel.predict() + * + * @param featuresBytes Serialized feature vector for data point + * @return predicted label + */ + def predictDecisionTreeModel( + model: DecisionTreeModel, + featuresBytes: Array[Byte]): Double = { + val features: Vector = deserializeDoubleVector(featuresBytes) + model.predict(features) + } + + /** + * Predict the labels of the given data points. + * This is a Java stub for python DecisionTreeModel.predict() + * + * @param dataJRDD A JavaRDD with serialized feature vectors + * @return JavaRDD of serialized predictions + */ + def predictDecisionTreeModel( + model: DecisionTreeModel, + dataJRDD: JavaRDD[Array[Byte]]): JavaRDD[Array[Byte]] = { + val data = dataJRDD.rdd.map(xBytes => deserializeDoubleVector(xBytes)) + model.predict(data).map(serializeDouble) + } + + /** * Java stub for mllib Statistics.corr(X: RDD[Vector], method: String). * Returns the correlation matrix serialized into a byte array understood by deserializers in * pyspark. @@ -597,4 +674,5 @@ class PythonMLLibAPI extends Serializable { val s = getSeedOrDefault(seed) RG.poissonVectorRDD(jsc.sc, mean, numRows, numCols, parts, s).map(serializeDoubleVector) } + } diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala index 5c65b537b6..fdad4f029a 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/configuration/Strategy.scala @@ -56,7 +56,8 @@ class Strategy ( if (algo == Classification) { require(numClassesForClassification >= 2) } - val isMulticlassClassification = numClassesForClassification > 2 + val isMulticlassClassification = + algo == Classification && numClassesForClassification > 2 val isMulticlassWithCategoricalFeatures = isMulticlassClassification && (categoricalFeaturesInfo.size > 0) diff --git a/mllib/src/test/scala/org/apache/spark/mllib/tree/DecisionTreeSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/tree/DecisionTreeSuite.scala index 546a132559..8665a00f3b 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/tree/DecisionTreeSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/tree/DecisionTreeSuite.scala @@ -48,7 +48,8 @@ class DecisionTreeSuite extends FunSuite with LocalSparkContext { requiredMSE: Double) { val predictions = input.map(x => model.predict(x.features)) val squaredError = predictions.zip(input).map { case (prediction, expected) => - (prediction - expected.label) * (prediction - expected.label) + val err = prediction - expected.label + err * err }.sum val mse = squaredError / input.length assert(mse <= requiredMSE) |