1.目標(biāo)：

多分類，計(jì)算混淆矩陣confusion_matrix，以及accuracy、precision、recall、f1-score分?jǐn)?shù)。

2.代碼：

• 1）使用sklearn計(jì)算并畫出混淆矩陣（confusion_matrix）；

• 2）使用sklearn計(jì)算accuracy（accuracy_score）；

• 3）使用sklearn計(jì)算多分類的precision、recall、f1-score分?jǐn)?shù)。以及計(jì)算每個(gè)類別的precision、recall、f1-score。

• precision：precision_score https://scikit-learn.org/stable/modules/generated/sklearn.metrics.precision_score.html

• recall：recall_score https://scikit-learn.org/stable/modules/generated/sklearn.metrics.recall_score.html

• F1 score：f1_score https://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html

• 默認(rèn)precision_score、recall_score、f1_score函數(shù)只能求出二分類問題的精準(zhǔn)率（average參數(shù)默認(rèn)為binary），不過如果要想能夠求出多分類問題的精準(zhǔn)率可以選擇合理average參數(shù)的值，比如：micro、macro等。

• 參數(shù)average有5個(gè)選項(xiàng)：{‘micro’微平均, ‘macro’宏平均, ‘samples’, ‘weighted’, ‘binary’}，默認(rèn)是default=’binary’二分類。

import seaborn as sns
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score, precision_score, f1_score, recall_score, classification_report

# y_true為真實(shí)值，y_pred為預(yù)測值（此處y_true和y_pred僅作舉例，隨便取的值，有0~4共5個(gè)類別。）
y_true = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4]
y_pred = [1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 0, 3, 3, 3, 0, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4]

# 1.計(jì)算混淆矩陣
cm = confusion_matrix(y_true, y_pred)
conf_matrix = pd.DataFrame(cm, index=['1','2','3','4','5'], columns=['1','2','3','4','5'])  #數(shù)據(jù)有5個(gè)類別
# 畫出混淆矩陣
fig, ax = plt.subplots(figsize=(4.5, 3.5))
sns.heatmap(conf_matrix, annot=True, annot_kws={"size": 14}, cmap="Blues")
plt.ylabel('True label', fontsize=14)
plt.xlabel('Predicted label', fontsize=14)
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.savefig('confusion.pdf', bbox_inches='tight')
plt.show()

# 2.計(jì)算accuracy
print('accuracy_score', accuracy_score(y_true, y_pred))

# 3.計(jì)算多分類的precision、recall、f1-score分?jǐn)?shù)
print('Micro precision', precision_score(y_true, y_pred, average='micro'))
print('Micro recall', recall_score(y_true, y_pred, average='micro'))
print('Micro f1-score', f1_score(y_true, y_pred, average='micro'))

print('Macro precision', precision_score(y_true, y_pred, average='macro'))
print('Macro recall', recall_score(y_true, y_pred, average='macro'))
print('Macro f1-score', f1_score(y_true, y_pred, average='macro'))

# 下面這個(gè)可以顯示出每個(gè)類別的precision、recall、f1-score。
print('classification_report\n',classification_report(y_true, y_pred))

3.結(jié)果：

參考：

https://blog.csdn.net/kan2281123066/article/details/103237273 利用sklearn 計(jì)算 precision、recall、F1 score

https://zhuanlan.zhihu.com/p/147663370 多分類模型Accuracy, Precision, Recall和F1-score的超級無敵深入探討

https://cloud.tencent.com/developer/article/1632611 機(jī)器學(xué)習(xí)入門 10-8 多分類問題中的混淆矩陣文章來源地址http://www.zghlxwxcb.cn/news/detail-463343.html