Spark中决策树源码分析

1.Example

使用Spark MLlib中决策树分类器API，训练出一个决策树模型，使用Python开发。

"""
Decision Tree Classification Example.
"""
from __future__ import print_function

from pyspark import SparkContext
from pyspark.mllib.tree import DecisionTree, DecisionTreeModel
from pyspark.mllib.util import MLUtils

if __name__ == "__main__":

    sc = SparkContext(appName="PythonDecisionTreeClassificationExample")

    # 加载和解析数据文件为RDD
    dataPath = "/home/zhb/Desktop/work/DecisionTreeShareProject/app/sample_libsvm_data.txt"
    print(dataPath)

    data = MLUtils.loadLibSVMFile(sc,dataPath)
    # 将数据集分割为训练数据集和测试数据集
    (trainingData,testData) = data.randomSplit([0.7,0.3])
    print("train data count: " + str(trainingData.count()))
    print("test data count : " + str(testData.count()))

    # 训练决策树分类器
    # categoricalFeaturesInfo 为空，表示所有的特征均为连续值
    model = DecisionTree.trainClassifier(trainingData, numClasses=2, categoricalFeaturesInfo={},
                                         impurity='gini', maxDepth=5, maxBins=32)

    # 测试数据集上预测
    predictions = model.predict(testData.map(lambda x: x.features))
    # 打包真实值与预测值
    labelsAndPredictions = testData.map(lambda lp: lp.label).zip(predictions)
    # 统计预测错误的样本的频率
    testErr = labelsAndPredictions.filter(lambda (v, p): v != p).count() / float(testData.count())
    print('Decision Tree Test Error = %5.3f%%'%(testErr*100))
    print("Decision Tree Learned classifiction tree model : ")
    print(model.toDebugString())

    # 保存和加载训练好的模型
    modelPath = "/home/zhb/Desktop/work/DecisionTreeShareProject/app/myDecisionTreeClassificationModel"
    model.save(sc, modelPath)
    sameModel = DecisionTreeModel.load(sc, modelPath)

2.决策树源码分析

决策树分类器API为DecisionTree.trainClassifier，进入源码分析。

源码文件所在路径为，spark-1.6/mllib/src/main/scala/org/apache/spark/mllib/tree/DecisionTree.scala。

  @Since("1.1.0")
  def trainClassifier(
      input: RDD[LabeledPoint],
      numClasses: Int,
      categoricalFeaturesInfo: Map[Int, Int],
      impurity: String,
      maxDepth: Int,
      maxBins: Int): DecisionTreeModel = {
    val impurityType = Impurities.fromString(impurity)
    train(input, Classification, impurityType, maxDepth, numClasses, maxBins, Sort,
      categoricalFeaturesInfo)
  }

训练出一个分类器，然后调用了train方法。

  @Since("1.0.0")
  def train(
      input: RDD[LabeledPoint],
      algo: Algo,
      impurity: Impurity,
      maxDepth: Int,
      numClasses: Int,
      maxBins: Int,
      quantileCalculationStrategy: QuantileStrategy,
      categoricalFeaturesInfo: Map[Int, Int]): DecisionTreeModel = {
    val strategy = new Strategy(algo, impurity, maxDepth, numClasses, maxBins,
      quantileCalculationStrategy, categoricalFeaturesInfo)
    new DecisionTree(strategy).run(input)
  }

train方法首先将模型类型（分类或者回归）、信息增益指标、决策树深度、分类数目、最大切分箱子数等参数封装为Strategy，然后新建一个DecisionTree对象，并调用run方法。

@Since("1.0.0")
class DecisionTree private[spark] (private val strategy: Strategy, private val seed: Int)
  extends Serializable with Logging {

  /**
   * @param strategy The configuration parameters for the tree algorithm which specify the type
   *                 of decision tree (classification or regression), feature type (continuous,
   *                 categorical), depth of the tree, quantile calculation strategy, etc.
   */
  @Since("1.0.0")
  def this(strategy: Strategy) = this(strategy, seed = 0)

  strategy.assertValid()

  /**
   * Method to train a decision tree model over an RDD
   *
   * @param input Training data: RDD of [[org.apache.spark.mllib.regression.LabeledPoint]].
   * @return DecisionTreeModel that can be used for prediction.
   */
  @Since("1.2.0")
  def run(input: RDD[LabeledPoint]): DecisionTreeModel = {
    val rf = new RandomForest(strategy, numTrees = 1, featureSubsetStrategy = "all", seed = seed)
    val rfModel = rf.run(input)
    rfModel.trees(0)
  }
}

run方法中首先新建一个RandomForest对象，将strategy、决策树数目设置为1，子集选择策略为"all"传递给RandomForest对象，然后调用RandomForest中的run方法，最后返回随机森林模型中的第一棵决策树。

也就是，决策树模型使用了随机森林模型进行训练，将决策树数目设置为1，然后将随机森林模型中的第一棵决策树作为结果，返回作为决策树训练模型。

3.随机森林源码分析

随机森林的源码文件所在路径为，spark-1.6/mllib/src/main/scala/org/apache/spark/mllib/tree/RandomForest.scala。

private class RandomForest (
    private val strategy: Strategy,
    private val numTrees: Int,
    featureSubsetStrategy: String,
    private val seed: Int)
  extends Serializable with Logging {

  strategy.assertValid()
  require(numTrees > 0, s"RandomForest requires numTrees > 0, but was given numTrees = $numTrees.")
  require(RandomForest.supportedFeatureSubsetStrategies.contains(featureSubsetStrategy)
    || Try(featureSubsetStrategy.toInt).filter(_ > 0).isSuccess
    || Try(featureSubsetStrategy.toDouble).filter(_ > 0).filter(_ <= 1.0).isSuccess,
    s"RandomForest given invalid featureSubsetStrategy: $featureSubsetStrategy." +
    s" Supported values: ${NewRFParams.supportedFeatureSubsetStrategies.mkString(", ")}," +
    s" (0.0-1.0], [1-n].")

  /**
   * Method to train a decision tree model over an RDD
   *
   * @param input Training data: RDD of [[org.apache.spark.mllib.regression.LabeledPoint]].
   * @return RandomForestModel that can be used for prediction.
   */
  def run(input: RDD[LabeledPoint]): RandomForestModel = {
    val trees: Array[NewDTModel] = NewRandomForest.run(input.map(_.asML), strategy, numTrees,
      featureSubsetStrategy, seed.toLong, None)
    new RandomForestModel(strategy.algo, trees.map(_.toOld))
  }

}

在该文件开头，通过"import org.apache.spark.ml.tree.impl.{RandomForest => NewRandomForest}"将ml中的RandomForest引入，重新命名为NewRandomForest。

在RandomForest.run方法中，首先新建NewRandomForest模型，并调用该类的run方法，然后将生成的trees作为新建RandomForestModel的入参。

NewRandomForest，源码文件所在路径为，spark-1.6/mllib/src/main/scala/org/apache/spark/ml/tree/impl/RandomForest.scala。

由于涉及代码量较大，因此无法将代码展开，run方法主要有如下调用。

run方法

--->1. val metadata = DecisionTreeMetadata.buildMetadata(retaggedInput, strategy, numTrees,featureSubsetStrategy) # 对输入数据建立元数据

--->2. val splits = findSplits(retaggedInput, metadata, seed) # 对元数据中的特征进行切分

    --->2.1 计算采样率，对输入样本进行采样

    --->2.2 findSplitsBySorting(sampledInput, metadata, continuousFeatures) # 对采样后的样本中的特征进行切分

        --->2.2.1 val thresholds = findSplitsForContinuousFeature(samples, metadata, idx) # 针对连续型特征

        --->2.2.2 val categories = extractMultiClassCategories(splitIndex + 1, featureArity) # 针对分类型特征，且特征无序

        --->2.2.3 Array.empty[Split] # 针对分类型特征，且特征有序，训练时直接构造即可

--->3. val treeInput = TreePoint.convertToTreeRDD(retaggedInput, splits, metadata) # 将输入数据转换为树形数据

    --->3.1 input.map { x => TreePoint.labeledPointToTreePoint(x, thresholds, featureArity) # 将LabeledPoint数据转换为TreePoint数据

    --->3.2 arr(featureIndex) = findBin(featureIndex, labeledPoint, featureArity(featureIndex), thresholds(featureIndex)) # 在(labeledPoint,feature)中找出一个离散值

--->4. val baggedInput = BaggedPoint.convertToBaggedRDD(treeInput, strategy.subsamplingRate, numTrees,withReplacement, seed) # 对输入数据进行采样

    --->4.1 convertToBaggedRDDSamplingWithReplacement(input, subsamplingRate, numSubsamples, seed) #有放回采样

    --->4.2 convertToBaggedRDDWithoutSampling(input) # 样本数为1，采样率为100%

    --->4.3 convertToBaggedRDDSamplingWithoutReplacement(input, subsamplingRate, numSubsamples, seed) # 无放回采样

--->5. val (nodesForGroup, treeToNodeToIndexInfo) = RandomForest.selectNodesToSplit(nodeQueue, maxMemoryUsage,metadata, rng) # 取得每棵树所有需要切分的结点

    --->5.1 val featureSubset: Option[Array[Int]] = if (metadata.subsamplingFeatures) { Some(SamplingUtils.reservoirSampleAndCount(Range(0, metadata.numFeatures).iterator, metadata.numFeaturesPerNode, rng.nextLong())._1)} # 如果需要子采样，选择特征子集

    --->5.2 val nodeMemUsage = RandomForest.aggregateSizeForNode(metadata, featureSubset) * 8L # 计算添加这个结点之后，是否有足够的内存

--->6. RandomForest.findBestSplits(baggedInput, metadata, topNodes, nodesForGroup, treeToNodeToIndexInfo, splits, nodeQueue, timer, nodeIdCache) # 找出最优切分点

    --->6.1 val (split: Split, stats: ImpurityStats) = binsToBestSplit(aggStats, splits, featuresForNode, nodes(nodeIndex)) #找出每个结点最好的切分

--->7. new DecisionTreeClassificationModel(uid, rootNode.toNode, numFeatures, strategy.getNumClasses) # 返回决策树分类模型

4.Reference

spark mllib中的随机森林算法，实现源码以及使用介绍

Spark MLlib - Decision Tree源码分析

Spark MLlib机器学习：算法、源码及实战详解