mean_tpr = 0.0 
  mean_fpr = np.linspace(0, 1, 100) 
  all_tpr = []
  
  y_target = np.r_[train_y,test_y]
  cv = StratifiedKFold(y_target, n_folds=6)
 
    #畫ROC曲線和計(jì)算AUC
    fpr, tpr, thresholds = roc_curve(test_y, predict,pos_label = 2)##指定正例標(biāo)簽，pos_label = ###########在數(shù)之聯(lián)的時(shí)候?qū)W到的，要制定正例
    
    mean_tpr += interp(mean_fpr, fpr, tpr)     #對(duì)mean_tpr在mean_fpr處進(jìn)行插值，通過scipy包調(diào)用interp()函數(shù) 
    mean_tpr[0] = 0.0                #初始處為0 
    roc_auc = auc(fpr, tpr) 
    #畫圖，只需要plt.plot(fpr,tpr),變量roc_auc只是記錄auc的值，通過auc()函數(shù)能計(jì)算出來 
    plt.plot(fpr, tpr, lw=1, label='ROC %s (area = %0.3f)' % (classifier, roc_auc)) 

# -*- coding: utf-8 -*- 
""" 
Created on Sun Apr 19 08:57:13 2015 
@author: shifeng 
""" 
print(__doc__) 
 
import numpy as np 
from scipy import interp 
import matplotlib.pyplot as plt 
 
from sklearn import svm, datasets 
from sklearn.metrics import roc_curve, auc 
from sklearn.cross_validation import StratifiedKFold 
 
############################################################################### 
# Data IO and generation,導(dǎo)入iris數(shù)據(jù)，做數(shù)據(jù)準(zhǔn)備 
 
# import some data to play with 
iris = datasets.load_iris() 
X = iris.data 
y = iris.target 
X, y = X[y != 2], y[y != 2]#去掉了label為2，label只能二分，才可以。 
n_samples, n_features = X.shape 
 
# Add noisy features 
random_state = np.random.RandomState(0) 
X = np.c_[X, random_state.randn(n_samples, 200 * n_features)] 
 
############################################################################### 
# Classification and ROC analysis 
#分類，做ROC分析 
 
# Run classifier with cross-validation and plot ROC curves 
#使用6折交叉驗(yàn)證，并且畫ROC曲線 
cv = StratifiedKFold(y, n_folds=6) 
classifier = svm.SVC(kernel='linear', probability=True, 
           random_state=random_state)#注意這里，probability=True,需要，不然預(yù)測的時(shí)候會(huì)出現(xiàn)異常。另外rbf核效果更好些。 
mean_tpr = 0.0 
mean_fpr = np.linspace(0, 1, 100) 
all_tpr = [] 
 
for i, (train, test) in enumerate(cv): 
  #通過訓(xùn)練數(shù)據(jù)，使用svm線性核建立模型，并對(duì)測試集進(jìn)行測試，求出預(yù)測得分 
  probas_ = classifier.fit(X[train], y[train]).predict_proba(X[test]) 
#  print set(y[train])           #set([0,1]) 即label有兩個(gè)類別 
#  print len(X[train]),len(X[test])    #訓(xùn)練集有84個(gè)，測試集有16個(gè) 
#  print "++",probas_           #predict_proba()函數(shù)輸出的是測試集在lael各類別上的置信度， 
#  #在哪個(gè)類別上的置信度高，則分為哪類 
  # Compute ROC curve and area the curve 
  #通過roc_curve()函數(shù)，求出fpr和tpr，以及閾值 
  fpr, tpr, thresholds = roc_curve(y[test], probas_[:, 1]) 
  mean_tpr += interp(mean_fpr, fpr, tpr)     #對(duì)mean_tpr在mean_fpr處進(jìn)行插值，通過scipy包調(diào)用interp()函數(shù) 
  mean_tpr[0] = 0.0                #初始處為0 
  roc_auc = auc(fpr, tpr) 
  #畫圖，只需要plt.plot(fpr,tpr),變量roc_auc只是記錄auc的值，通過auc()函數(shù)能計(jì)算出來 
  plt.plot(fpr, tpr, lw=1, label='ROC fold %d (area = %0.2f)' % (i, roc_auc)) 
 
#畫對(duì)角線 
plt.plot([0, 1], [0, 1], '--', color=(0.6, 0.6, 0.6), label='Luck') 
 
mean_tpr /= len(cv)           #在mean_fpr100個(gè)點(diǎn)，每個(gè)點(diǎn)處插值插值多次取平均 
mean_tpr[-1] = 1.0           #坐標(biāo)最后一個(gè)點(diǎn)為（1,1） 
mean_auc = auc(mean_fpr, mean_tpr)   #計(jì)算平均AUC值 
#畫平均ROC曲線 
#print mean_fpr,len(mean_fpr) 
#print mean_tpr 
plt.plot(mean_fpr, mean_tpr, 'k--', 
     label='Mean ROC (area = %0.2f)' % mean_auc, lw=2) 
 
plt.xlim([-0.05, 1.05]) 
plt.ylim([-0.05, 1.05]) 
plt.xlabel('False Positive Rate') 
plt.ylabel('True Positive Rate') 
plt.title('Receiver operating characteristic example') 
plt.legend(loc="lower right") 
plt.show() 

import org.apache.spark.ml.Pipeline
import org.apache.spark.ml.feature.{StringIndexer, OneHotEncoder}
import org.apache.spark.ml.feature.VectorAssembler
import ml.dmlc.xgboost4j.scala.spark.{XGBoostEstimator, XGBoostClassificationModel}
import org.apache.spark.ml.evaluation.BinaryClassificationEvaluator
import org.apache.spark.ml.tuning.{ParamGridBuilder, CrossValidator}
import org.apache.spark.ml.PipelineModel
 
val data = (spark.read.format("csv")
 .option("sep", ",")
 .option("inferSchema", "true")
 .option("header", "true")
 .load("/Affairs.csv"))
 
data.createOrReplaceTempView("res1")
val affairs = "case when affairs>0 then 1 else 0 end as affairs,"
val df = (spark.sql("select " + affairs +
 "gender,age,yearsmarried,children,religiousness,education,occupation,rating" +
 " from res1 "))
 
val categoricals = df.dtypes.filter(_._2 == "StringType") map (_._1)
val indexers = categoricals.map(
 c => new StringIndexer().setInputCol(c).setOutputCol(s"${c}_idx")
)
 
val encoders = categoricals.map(
 c => new OneHotEncoder().setInputCol(s"${c}_idx").setOutputCol(s"${c}_enc").setDropLast(false)
)
 
val colArray_enc = categoricals.map(x => x + "_enc")
val colArray_numeric = df.dtypes.filter(_._2 != "StringType") map (_._1)
val final_colArray = (colArray_numeric ++ colArray_enc).filter(!_.contains("affairs"))
val vectorAssembler = new VectorAssembler().setInputCols(final_colArray).setOutputCol("features")
 
/*
val pipeline = new Pipeline().setStages(indexers ++ encoders ++ Array(vectorAssembler))
pipeline.fit(df).transform(df)
*/
 
///
// Create an XGBoost Classifier 
val xgb = new XGBoostEstimator(Map("num_class" -> 2, "num_rounds" -> 5, "objective" -> "binary:logistic", "booster" -> "gbtree")).setLabelCol("affairs").setFeaturesCol("features")
 
// XGBoost paramater grid
val xgbParamGrid = (new ParamGridBuilder()
  .addGrid(xgb.round, Array(10))
  .addGrid(xgb.maxDepth, Array(10,20))
  .addGrid(xgb.minChildWeight, Array(0.1))
  .addGrid(xgb.gamma, Array(0.1))
  .addGrid(xgb.subSample, Array(0.8))
  .addGrid(xgb.colSampleByTree, Array(0.90))
  .addGrid(xgb.alpha, Array(0.0))
  .addGrid(xgb.lambda, Array(0.6))
  .addGrid(xgb.scalePosWeight, Array(0.1))
  .addGrid(xgb.eta, Array(0.4))
  .addGrid(xgb.boosterType, Array("gbtree"))
  .addGrid(xgb.objective, Array("binary:logistic")) 
  .build())
 
// Create the XGBoost pipeline
val pipeline = new Pipeline().setStages(indexers ++ encoders ++ Array(vectorAssembler, xgb))
 
// Setup the binary classifier evaluator
val evaluator = (new BinaryClassificationEvaluator()
  .setLabelCol("affairs")
  .setRawPredictionCol("prediction")
  .setMetricName("areaUnderROC"))
 
// Create the Cross Validation pipeline, using XGBoost as the estimator, the
// Binary Classification evaluator, and xgbParamGrid for hyperparameters
val cv = (new CrossValidator()
  .setEstimator(pipeline)
  .setEvaluator(evaluator)
  .setEstimatorParamMaps(xgbParamGrid)
  .setNumFolds(3)
  .setSeed(0))
 
 // Create the model by fitting the training data
val xgbModel = cv.fit(df)
 
 // Test the data by scoring the model
val results = xgbModel.transform(df)
 
// Print out a copy of the parameters used by XGBoost, attention pipeline
(xgbModel.bestModel.asInstanceOf[PipelineModel]
 .stages(5).asInstanceOf[XGBoostClassificationModel]
 .extractParamMap().toSeq.foreach(println))
results.select("affairs","prediction").show
 
println("---Confusion Matrix------")
results.stat.crosstab("affairs","prediction").show()
 
// What was the overall accuracy of the model, using AUC
val auc = evaluator.evaluate(results)
println("----AUC--------")
println("auc="+auc)