使用sklearn画二分类模型ROC曲线，PR曲线

1. 在digits数据集上训练模型

import matplotlib.pyplot as plt
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import plot_roc_curve
from sklearn.datasets import load_breast_cancer
from sklearn import datasets
from sklearn.model_selection import train_test_split
from sklearn import metrics

digits = datasets.load_digits()
X = digits.data
y = digits.target
y = [1 if label >= 5 else 0 for label in y]

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)

# 创建模型
svc_clf = SVC(probability=True)
lr_clf = LogisticRegression(solver='saga', max_iter=100)
dt_clf = DecisionTreeClassifier(min_samples_leaf=5, max_depth=8)
knn_clf = KNeighborsClassifier()

# 训练模型
svc_clf.fit(X_train, y_train)
lr_clf.fit(X_train, y_train)
dt_clf.fit(X_train, y_train)
knn_clf.fit(X_train, y_train)

2. 使用plot_roc_curve函数绘制ROC曲线

#创建画布
fig, ax = plt.subplots()

# svc_roc = plot_roc_curve(svc_clf, X_test, y_test, ax=ax)
lr_clf_roc = plot_roc_curve(lr_clf, X_test, y_test, ax=ax)
dt_clf_roc = plot_roc_curve(dt_clf, X_test, y_test, ax=ax)
# knn_clf_roc = plot_roc_curve(knn_clf, X_test, y_test, ax=ax)

# 参照线
ax.plot([0, 1], [0, 1], linestyle='--', color='r')

image.png

3. 使用roc_curve函数绘制ROC曲线

from sklearn import svm, datasets
from sklearn import metrics
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.datasets import load_breast_cancer
import matplotlib.pyplot as plt
import numpy as np
from sklearn.metrics import roc_curve,auc,roc_auc_score

# 模型预测
y_pred_svc = svc_clf.predict_proba(X_test)[:,1]
y_pred_lr = lr_clf.predict_proba(X_test)[:,1]
y_pred_dt = dt_clf.predict_proba(X_test)[:,1]
y_pred_knn = knn_clf.predict_proba(X_test)[:,1]
y_pred_rand = np.random.rand(len(X_test)) # 随机生成概率

fpr_svc,tpr_svc,thres_svc = roc_curve(y_test,y_pred_svc)
fpr_lr,tpr_lr,thres_lr = roc_curve(y_test,y_pred_lr)
fpr_dt,tpr_dt,thres_dt = roc_curve(y_test,y_pred_dt)
fpr_knn,tpr_knn,thres_knn = roc_curve(y_test,y_pred_knn)
fpr_rand,tpr_rand,thres_rand = roc_curve(y_test,y_pred_rand)

print("SVC的AUC为:",auc(fpr_svc,tpr_svc))
print("LogitReg的AUC为:",auc(fpr_lr,tpr_lr))
print("DecisionTree的AUC为:",auc(fpr_dt,tpr_dt))
print("kNN的AUC为:",auc(fpr_knn,tpr_knn))
print("随机的AUC为:",auc(fpr_rand,tpr_rand))

#创建画布
fig,ax = plt.subplots()

#自定义标签名称label=''
# ax.plot(fpr_svc,tpr_svc,linewidth=2,
#         label='Random (AUC={})'.format(str(round(auc(fpr_svc,tpr_svc),3))))
ax.plot(fpr_lr,tpr_lr,linewidth=2,
        label='Logistic Regression (AUC={})'.format(str(round(auc(fpr_lr,tpr_lr),3))))
ax.plot(fpr_dt,tpr_dt,linewidth=2,
        label='Decision Tree (AUC={})'.format(str(round(auc(fpr_dt,tpr_dt),3))))
# ax.plot(fpr_knn,tpr_knn,linewidth=2,
#         label='K Nearest Neibor (AUC={})'.format(str(round(auc(fpr_knn,tpr_knn),3))))
# ax.plot(fpr_rand,tpr_rand,linewidth=2,
#         label='Random (AUC={})'.format(str(round(auc(fpr_rand,tpr_rand),3))))
#绘制对角线
ax.plot([0,1],[0,1],linestyle='--',color='grey')

#调整字体大小
plt.legend(fontsize=12)

image.png

4. 使用precision_recall_curve函数绘制PR曲线

from sklearn.metrics import precision_recall_curve

precision, recall, threshold = precision_recall_curve(y_test, y_pred_lr, pos_label=1)
fig = plt.figure()
plt.plot(precision, recall, label='Logistic')

plt.xlabel('Recall')
plt.ylabel('Precision')
plt.legend()

image.png

注意事项

使用roc_auc_score()计算AUC的时候，传入的第一个参数应该是预测的真实标签，第二个参数应该是模型预测为“真(1)”的概率而不是模型预测的“0-1标签”。如果传入后者，会造成比实际AUC值偏低的情况。