利用鸢尾花数据集绘制P-R曲线

Posted 2020-12-23 cxq1126

 1 #利用鸢尾花数据集绘制P-R曲线
 2 print(__doc__)      #打印注释
 3  
 4 import matplotlib.pyplot as plt
 5 import numpy as np
 6 from sklearn import svm, datasets
 7 from sklearn.metrics import precision_recall_curve
 8 from sklearn.metrics import average_precision_score
 9 from sklearn.preprocessing import label_binarize
10 from sklearn.multiclass import OneVsRestClassifier  #一对其余（每次将一个类作为正类，剩下的类作为负类）
11  
12 from sklearn.cross_validation import train_test_split  #适用于anaconda 3.6及以前版本
13 #from sklearn.model_selection import train_test_split#适用于anaconda 3.7
14  
15 #以iris数据为例，画出P-R曲线
16 iris = datasets.load_iris()
17 X = iris.data    #150*4
18 y = iris.target  #150*1
19  
20 # 标签二值化,将三个类转为001, 010, 100的格式.因为这是个多类分类问题，后面将要采用
21 #OneVsRestClassifier策略转为二类分类问题
22 y = label_binarize(y, classes=[0, 1, 2])    #将150*1转化成150*3
23 n_classes = y.shape[1]                      #列的个数，等于3
24 print (y)
25  
26 # 增加了800维的噪声特征
27 random_state = np.random.RandomState(0)
28 n_samples, n_features = X.shape
29  
30 X = np.c_[X, random_state.randn(n_samples, 200 * n_features)]   #行不变，只增加了列，150*804
31  
32 # 训练集和测试集拆分，比例为0.5
33 X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.5, random_state=random_state) #随机数，填0或不填，每次都会不一样
34  
35 # 一对其余，转换成两类，构建新的分类器
36 classifier = OneVsRestClassifier(svm.SVC(kernel=‘linear‘, probability=True, random_state=random_state))
37 #训练集送给fit函数进行拟合训练，训练完后将测试集的样本特征注入，得到测试集中每个样本预测的分数
38 y_score = classifier.fit(X_train, y_train).decision_function(X_test)
39  
40 # Compute Precision-Recall and plot curve  
41 #下面的下划线是返回的阈值。作为一个名称：此时“_”作为临时性的名称使用。
42 #表示分配了一个特定的名称，但是并不会在后面再次用到该名称。
43 precision = dict()
44 recall = dict()
45 average_precision = dict()
46 for i in range(n_classes):
47     #对于每一类，计算精确率和召回率的序列（:表示所有行，i表示第i列）
48     precision[i], recall[i], _ = precision_recall_curve(y_test[:, i],  y_score[:, i]) 
49     average_precision[i] = average_precision_score(y_test[:, i], y_score[:, i])#切片，第i个类的分类结果性能
50  
51 # Compute micro-average curve and area. ravel()将多维数组降为一维
52 precision["micro"], recall["micro"], _ = precision_recall_curve(y_test.ravel(),  y_score.ravel())
53 average_precision["micro"] = average_precision_score(y_test, y_score, average="micro") #This score corresponds to the area under the precision-recall curve.
54  
55 # Plot Precision-Recall curve for each class
56 plt.clf()#clf 函数用于清除当前图像窗口
57 plt.plot(recall["micro"], precision["micro"],
58          label=‘micro-average Precision-recall curve (area = {0:0.2f})‘.format(average_precision["micro"]))
59 for i in range(n_classes):
60     plt.plot(recall[i], precision[i],
61              label=‘Precision-recall curve of class {0} (area = {1:0.2f})‘.format(i, average_precision[i]))
62  
63 plt.xlim([0.0, 1.0])
64 plt.ylim([0.0, 1.05]) #xlim、ylim：分别设置X、Y轴的显示范围。
65 plt.xlabel(‘Recall‘, fontsize=16)
66 plt.ylabel(‘Precision‘,fontsize=16)
67 plt.title(‘Extension of Precision-Recall curve to multi-class‘,fontsize=16)
68 plt.legend(loc="lower right")#legend 是用于设置图例的函数
69 plt.show()

运行结果如下：