Spark中组件Mllib的学习39之梯度提升树（GBT）用于分类*

更多代码请见：https://github.com/xubo245/SparkLearning

Spark中组件Mllib的学习之分类篇

1解释

（1）GBDT基本概念

用ID3算法和C4.5算法学习得到的决策树，有可能导致模型过拟合，通常使用剪枝算法来解决。随着集成学习的发展，出现了比较典型的迭代决策树GBDT和随机森林RF，即将多棵单决策树进行模型组合，形成多决策树，可以看成Treelink。

迭代决策树有以下名称：

GBDT（Gradient Boost Decision Tree）渐进梯度决策树

GBRT（Gradient Boost Regression Tree）渐进梯度回归树

MART（Multiple Additive Regression Tree）多决策回归树

请参考【4】、【5】、【6】

（2）梯度提升树用于分类

【1】中给的方法有问题，需要进行修改，另外：

boostingStrategy.treeStrategy.categoricalFeaturesInfo = Map[Int, Int]()

没办法设置

boostingStrategy.treeStrategy.setCategoricalFeaturesInfo( java.util.Map[Integer, Integer]())

也不行，只好注释掉了。

2.代码：

/**
* @author xubo
*         ref:Spark MlLib机器学习实战
*         more code:https://github.com/xubo245/SparkLearning
*         more blog:http://blog.csdn.net/xubo245
*/
package org.apache.spark.mllib.learning.classification

import java.text.SimpleDateFormat
import java.util.Date

import org.apache.spark.mllib.tree.DecisionTree
import org.apache.spark.mllib.util.MLUtils
import org.apache.spark.{SparkConf, SparkContext}
import org.apache.spark.mllib.tree.GradientBoostedTrees
import org.apache.spark.mllib.tree.configuration.BoostingStrategy
import org.apache.spark.mllib.tree.model.{DecisionTreeModel, GradientBoostedTreesModel}
import org.apache.spark.mllib.util.MLUtils
import java.util.Map

/**
* Created by xubo on 2016/5/23.
*/
object GBTs1 {
def main(args: Array[String]) {
val conf = new SparkConf().setMaster("local[4]").setAppName(this.getClass().getSimpleName().filter(!_.equals('$')))
val sc = new SparkContext(conf)

// Load and parse the data file.
val data = MLUtils.loadLibSVMFile(sc, "file/data/mllib/input/classification/sample_libsvm_data.txt")

// Split the data into training and test sets (30% held out for testing)
val splits = data.randomSplit(Array(0.7, 0.3))
val (trainingData, testData) = (splits(0), splits(1))

// Train a GradientBoostedTrees model.
//  The defaultParams for Classification use LogLoss by default.
var boostingStrategy = BoostingStrategy.defaultParams("Classification")
boostingStrategy.setNumIterations(3)
boostingStrategy.treeStrategy.setNumClasses(2)
boostingStrategy.treeStrategy.setMaxDepth(5)
//   boostingStrategy.treeStrategy.setCategoricalFeaturesInfo( java.util.Map[Integer, Integer]())

//error
//    boostingStrategy.numIterations = 3 // Note: Use more iterations in practice.
//    boostingStrategy.treeStrategy.numClasses = 2
//    boostingStrategy.treeStrategy.maxDepth = 5
//    //  Empty categoricalFeaturesInfo indicates all features are continuous.
//        boostingStrategy.treeStrategy.categoricalFeaturesInfo = Map[Int, Int]()

val model = GradientBoostedTrees.train(trainingData, boostingStrategy)

// Evaluate model on test instances and compute test error
val labelAndPreds = testData.map { point =>
val prediction = model.predict(point.features)
(point.label, prediction)
}
val testErr = labelAndPreds.filter(r => r._1 != r._2).count.toDouble / testData.count()
println("Test Error = " + testErr)
println("Learned classification GBT model:\n" + model.toDebugString)

println("data.count:" + data.count())
println("trainingData.count:" + trainingData.count())
println("testData.count:" + testData.count())
println("model.algo:" + model.algo)
println("model.trees:" + model.trees)
println("model.treeWeights:" + model.treeWeights)

println("labelAndPreds")
labelAndPreds.take(10).foreach(println)

//     Save and load model
//    val iString = new SimpleDateFormat("yyyyMMddHHmmssSSS").format(new Date())
//    val path = "file/data/mllib/output/classification/GradientBoostedTreesModel" + iString + "/result"
//    model.save(sc, path)
//    val sameModel = DecisionTreeModel.load(sc, path)
//    println(sameModel.algo)
sc.stop
}
}

3.结果：

Test Error = 0.034482758620689655
Learned classification GBT model:
TreeEnsembleModel classifier with 3 trees

Tree 0:
If (feature 406 <= 72.0)
If (feature 100 <= 165.0)
Predict: -1.0
Else (feature 100 > 165.0)
Predict: 1.0
Else (feature 406 > 72.0)
Predict: 1.0
Tree 1:
If (feature 433 <= 0.0)
If (feature 351 <= 251.0)
If (feature 183 <= 228.0)
Predict: -0.4768116880884702
Else (feature 183 > 228.0)
Predict: -0.4768116880884703
Else (feature 351 > 251.0)
Predict: 0.4768116880884694
Else (feature 433 > 0.0)
Predict: 0.47681168808847
Tree 2:
If (feature 434 <= 0.0)
If (feature 435 <= 0.0)
If (feature 241 <= 19.0)
Predict: -0.4381935810427206
Else (feature 241 > 19.0)
If (feature 215 <= 40.0)
Predict: -0.4381935810427206
Else (feature 215 > 40.0)
Predict: -0.43819358104272066
Else (feature 435 > 0.0)
Predict: 0.43819358104271977
Else (feature 434 > 0.0)
If (feature 123 <= 0.0)
If (feature 407 <= 218.0)
Predict: 0.4381935810427206
Else (feature 407 > 218.0)
Predict: 0.43819358104272066
Else (feature 123 > 0.0)
Predict: 0.43819358104272155

data.count:100
trainingData.count:71
testData.count:29
model.algo:Classification
model.trees:[Lorg.apache.spark.mllib.tree.model.DecisionTreeModel;@4d65c183
model.treeWeights:[D@5dbbb110
labelAndPreds
(1.0,1.0)
(0.0,0.0)
(0.0,0.0)
(0.0,0.0)
(1.0,1.0)
(1.0,1.0)
(1.0,1.0)
(1.0,1.0)
(0.0,0.0)
(0.0,0.0)

参考

【1】http://spark.apache.org/docs/1.5.2/mllib-guide.html

【2】http://spark.apache.org/docs/1.5.2/programming-guide.html

【3】https://github.com/xubo245/SparkLearning

【4】《统计学习方法》

【5】http://my.oschina.net/keyven/blog/615436

【6】http://blog.csdn.net/w28971023/article/details/8240756

【7】http://spark.apache.org/docs/1.5.2/mllib-ensembles.html