# -*-coding=utf-8 -*-

from pyspark import SparkConf, SparkContext

sc = SparkContext('local')

from pyspark.mllib.tree import GradientBoostedTrees, GradientBoostedTreesModel

from pyspark.mllib.util import MLUtils

# Load and parse the data file.

data = MLUtils.loadLibSVMFile(sc, "data/mllib/sample_libsvm_data.txt")

'''

每一行使用以下格式表示一个标记的稀疏特征向量

label index1:value1 index2:value2 ...

tempFile.write(b"+1 1:1.0 3:2.0 5:3.0\\n-1\\n-1 2:4.0 4:5.0 6:6.0")

>>> tempFile.flush()

>>> examples = MLUtils.loadLibSVMFile(sc, tempFile.name).collect()

>>> tempFile.close()

>>> examples[0]

LabeledPoint(1.0, (6,[0,2,4],[1.0,2.0,3.0]))

>>> examples[1]

LabeledPoint(-1.0, (6,[],[]))

>>> examples[2]

LabeledPoint(-1.0, (6,[1,3,5],[4.0,5.0,6.0]))

'''

# Split the data into training and test sets (30% held out for testing)  分割数据集，留30%作为测试集

(trainingData, testData) = data.randomSplit([0.7, 0.3])

# Train a GradientBoostedTrees model. 训练决策树模型

#  Notes: (a) Empty categoricalFeaturesInfo indicates all features are continuous. 空的categoricalFeaturesInfo意味着所有的特征都是连续的

#         (b) Use more iterations in practice. 在实践中使用更多的迭代步数

model = GradientBoostedTrees.trainClassifier(trainingData,

                                             categoricalFeaturesInfo={}, numIterations=30)

# Evaluate model on test instances and compute test error 评估模型

predictions = model.predict(testData.map(lambda x: x.features))

labelsAndPredictions = testData.map(lambda lp: lp.label).zip(predictions)

testErr = labelsAndPredictions.filter(

    lambda lp: lp[0] != lp[1]).count() / float(testData.count())

print('Test Error = ' + str(testErr)) #Test Error = 0.0

print('Learned classification GBT model:')

print(model.toDebugString())

'''

TreeEnsembleModel classifier with 30 trees

  Tree 0:

    If (feature 434 <= 0.0)

     If (feature 100 <= 165.0)

      Predict: -1.0

     Else (feature 100 > 165.0)

      Predict: 1.0

    Else (feature 434 > 0.0)

     Predict: 1.0

  Tree 1:

    If (feature 490 <= 0.0)

     If (feature 549 <= 253.0)

      If (feature 184 <= 0.0)

       Predict: -0.4768116880884702

      Else (feature 184 > 0.0)

       Predict: -0.47681168808847024

     Else (feature 549 > 253.0)

      Predict: 0.4768116880884694

    Else (feature 490 > 0.0)

     If (feature 215 <= 251.0)

      Predict: 0.4768116880884701

     Else (feature 215 > 251.0)

      Predict: 0.4768116880884712

  ...

  Tree 29:

    If (feature 434 <= 0.0)

     If (feature 209 <= 4.0)

      Predict: 0.1335953290513215

     Else (feature 209 > 4.0)

      If (feature 372 <= 84.0)

       Predict: -0.13359532905132146

      Else (feature 372 > 84.0)

       Predict: -0.1335953290513215

    Else (feature 434 > 0.0)

     Predict: 0.13359532905132146

'''

# Save and load model

model.save(sc, "myGradientBoostingClassificationModel")

sameModel = GradientBoostedTreesModel.load(sc,"myGradientBoostingClassificationModel")

print sameModel.predict(data.collect()[0].features) #0.0
返回目录
梯度提升树

【Spark机器学习速成宝典】模型篇07梯度提升树【Gradient-Boosted Trees】（Python版）

目录

梯度提升树原理

梯度提升树代码(Spark Python)
