利用scikitlearn画ROC曲线实例

一个完整的数据挖掘模型，最后都要进行模型评估，对于二分类来说，AUC，ROC这两个指标用到最多，所以 利用sklearn里面相应的函数进行模块搭建。

具体实现的代码可以参照下面博友的代码，评估svm的分类指标。注意里面的一些细节需要注意，一个是调用roc_curve 方法时，指明目标标签，否则会报错。

具体是这个参数的设置pos_label ，以前在unionbigdata实习时学到的。

重点是以下的代码需要根据实际改写：

 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)

#画ROC曲线和计算AUC
   fpr, tpr, thresholds = roc_curve(test_y, predict,pos_label = 2)##指定正例标签，pos_label = ###########在数之联的时候学到的，要制定正例

mean_tpr += interp(mean_fpr, fpr, tpr)     #对mean_tpr在mean_fpr处进行插值，通过scipy包调用interp()函数
   mean_tpr[0] = 0.0                #初始处为0
   roc_auc = auc(fpr, tpr)
   #画图，只需要plt.plot(fpr,tpr),变量roc_auc只是记录auc的值，通过auc()函数能计算出来
   plt.plot(fpr, tpr, lw=1, label='ROC ％s (area = ％0.3f)' ％ (classifier, roc_auc))

然后是博友的参考代码：

# -*- 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,导入iris数据，做数据准备

# 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为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
#分类，做ROC分析

# Run classifier with cross-validation and plot ROC curves
#使用6折交叉验证，并且画ROC曲线
cv = StratifiedKFold(y, n_folds=6)
classifier = svm.SVC(kernel='linear', probability=True,
          random_state=random_state)#注意这里，probability=True,需要，不然预测的时候会出现异常。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):
 #通过训练数据，使用svm线性核建立模型，并对测试集进行测试，求出预测得分
 probas_ = classifier.fit(X[train], y[train]).predict_proba(X[test])
#  print set(y[train])           #set([0,1]) 即label有两个类别
#  print len(X[train]),len(X[test])    #训练集有84个，测试集有16个
#  print "++",probas_           #predict_proba()函数输出的是测试集在lael各类别上的置信度，
#  #在哪个类别上的置信度高，则分为哪类
 # Compute ROC curve and area the curve
 #通过roc_curve()函数，求出fpr和tpr，以及阈值
 fpr, tpr, thresholds = roc_curve(y[test], probas_[:, 1])
 mean_tpr += interp(mean_fpr, fpr, tpr)     #对mean_tpr在mean_fpr处进行插值，通过scipy包调用interp()函数
 mean_tpr[0] = 0.0                #初始处为0
 roc_auc = auc(fpr, tpr)
 #画图，只需要plt.plot(fpr,tpr),变量roc_auc只是记录auc的值，通过auc()函数能计算出来
 plt.plot(fpr, tpr, lw=1, label='ROC fold ％d (area = ％0.2f)' ％ (i, roc_auc))

#画对角线
plt.plot([0, 1], [0, 1], '--', color=(0.6, 0.6, 0.6), label='Luck')

mean_tpr /= len(cv)           #在mean_fpr100个点，每个点处插值插值多次取平均
mean_tpr[-1] = 1.0           #坐标最后一个点为（1,1）
mean_auc = auc(mean_fpr, mean_tpr)   #计算平均AUC值
#画平均ROC曲线
#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()

补充知识：批量进行One-hot-encoder且进行特征字段拼接，并完成模型训练demo

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)

来源：https://blog.csdn.net/u010159842/article/details/53082003