Skip to main content
 首页 » 编程设计

python之具有显示标准偏差的 n 重交叉验证的精确召回曲线

2024年11月24日8qlqwjy

我想生成一 strip 有 5 折交叉验证的 Precision-Recall 曲线,显示标准偏差,如 example ROC curve code here 中所示.

下面的代码(改编自 How to Plot PR-Curve Over 10 folds of Cross Validation in Scikit-Learn)给出了每次交叉验证的 PR 曲线以及平均 PR 曲线。我还想用灰色显示平均 PR 曲线上下一个标准偏差的区域。但它给出了以下错误(代码下方链接中的详细信息):

ValueError:操作数无法与形状 (91,) (​​78,) 一起广播

import matplotlib.pyplot as plt 
import numpy 
from sklearn.datasets import make_blobs 
from sklearn.metrics import precision_recall_curve, auc 
from sklearn.model_selection import KFold 
from sklearn.svm import SVC 
 
 
X, y = make_blobs(n_samples=500, n_features=2, centers=2, cluster_std=10.0, 
    random_state=10) 
 
k_fold = KFold(n_splits=5, shuffle=True, random_state=10) 
predictor = SVC(kernel='linear', C=1.0, probability=True, random_state=10) 
 
y_real = [] 
y_proba = [] 
 
precisions, recalls = [], [] 
 
for i, (train_index, test_index) in enumerate(k_fold.split(X)): 
    Xtrain, Xtest = X[train_index], X[test_index] 
    ytrain, ytest = y[train_index], y[test_index] 
    predictor.fit(Xtrain, ytrain) 
    pred_proba = predictor.predict_proba(Xtest) 
    precision, recall, _ = precision_recall_curve(ytest, pred_proba[:,1]) 
    lab = 'Fold %d AUC=%.4f' % (i+1, auc(recall, precision)) 
    plt.plot(recall, precision, alpha=0.3, label=lab) 
    y_real.append(ytest) 
    y_proba.append(pred_proba[:,1]) 
    precisions.append(precision) 
    recalls.append(recall) 
 
y_real = numpy.concatenate(y_real) 
y_proba = numpy.concatenate(y_proba) 
precision, recall, _ = precision_recall_curve(y_real, y_proba) 
lab = 'Overall AUC=%.4f' % (auc(recall, precision)) 
 
plt.plot(recall, precision, lw=2,color='red', label=lab) 
 
std_precision = np.std(precisions, axis=0) 
tprs_upper = np.minimum(precisions[median] + std_precision, 1) 
tprs_lower = np.maximum(precisions[median] - std_precision, 0) 
plt.fill_between(recall_overall, upper_precision, lower_precision, alpha=0.5, linewidth=0, color='grey') 
 

Error reported and Plot generated

能否请您建议我如何添加到以下代码以显示平均 PR 曲线周围的一个标准差?

请您参考如下方法:

我有一个可行的解决方案,但如果有人可以评论它是否在做正确的事情,那将会很有帮助。

import matplotlib.pyplot as plt 
import numpy as np 
from sklearn.datasets import make_blobs 
from sklearn.metrics import precision_recall_curve, auc 
from sklearn.model_selection import KFold 
from sklearn.svm import SVC 
from numpy import interp 
 
X, y = make_blobs(n_samples=500, n_features=2, centers=2, cluster_std=10.0, 
    random_state=10) 
 
k_fold = KFold(n_splits=5, shuffle=True, random_state=10) 
predictor = SVC(kernel='linear', C=1.0, probability=True, random_state=10) 
 
y_real = [] 
y_proba = [] 
 
precision_array = [] 
threshold_array=[] 
recall_array = np.linspace(0, 1, 100) 
 
for i, (train_index, test_index) in enumerate(k_fold.split(X)): 
    Xtrain, Xtest = X[train_index], X[test_index] 
    ytrain, ytest = y[train_index], y[test_index] 
    predictor.fit(Xtrain, ytrain) 
    pred_proba = predictor.predict_proba(Xtest) 
    precision_fold, recall_fold, thresh = precision_recall_curve(ytest, pred_proba[:,1]) 
    precision_fold, recall_fold, thresh = precision_fold[::-1], recall_fold[::-1], thresh[::-1]  # reverse order of results 
    thresh = np.insert(thresh, 0, 1.0) 
    precision_array = interp(recall_array, recall_fold, precision_fold) 
    threshold_array = interp(recall_array, recall_fold, thresh) 
    pr_auc = auc(recall_array, precision_array) 
 
    lab_fold = 'Fold %d AUC=%.4f' % (i+1, pr_auc) 
    plt.plot(recall_fold, precision_fold, alpha=0.3, label=lab_fold) 
    y_real.append(ytest) 
    y_proba.append(pred_proba[:,1]) 
 
y_real = numpy.concatenate(y_real) 
y_proba = numpy.concatenate(y_proba) 
precision, recall, _ = precision_recall_curve(y_real, y_proba) 
lab = 'Overall AUC=%.4f' % (auc(recall, precision)) 
 
plt.plot(recall, precision, lw=2,color='red', label=lab) 
 
plt.legend(loc='lower left', fontsize='small') 
 
mean_precision = np.mean(precision_array) 
std_precision = np.std(precision_array) 
plt.fill_between(recall, precision + std_precision, precision - std_precision, alpha=0.3, linewidth=0, color='grey') 
plt.show()