python：使用sklearn 计算 precision、recall、F1 score（多分类）

参考：

0. 评估指标

1. 使用sklearn计算accuracy（accuracy_score）、混淆矩阵（confusion_matrix）

2. 使用sklearn计算精确率precision、召回率recall、F1 score

参考：

https://blog.csdn.net/kan2281123066/article/details/103237273 代码：利用sklearn 计算 precision、recall、F1 score

https://blog.csdn.net/blythe0107/article/details/75003890 代码：sklearn的precision_score, recall_score, f1_score使用

https://blog.csdn.net/Urbanears/article/details/105033731 二分类和多分类问题下的评价指标详析（Precision, Recall, F1，Micro，Macro）

https://blog.csdn.net/wf592523813/article/details/95202448 二分类和多分类问题的评价指标总结

https://zhuanlan.zhihu.com/p/147663370 知乎：多分类模型Accuracy, Precision, Recall和F1-score的超级无敌深入探讨

https://zhuanlan.zhihu.com/p/59862986 知乎：详解sklearn的多分类模型评价指标

https://www.zhihu.com/question/51470349 知乎：对多分类数据的模型比较选择，应该参考什么指标？

0. 评估指标

精确度：precision，正确预测为正的，占全部预测为正的比例，TP / (TP+FP)
召回率：recall，正确预测为正的，占全部实际为正的比例，TP / (TP+FN)
F1-score：精确率和召回率的调和平均数，2 * precision*recall / (precision+recall)

1. 使用sklearn计算accuracy（accuracy_score）、混淆矩阵（confusion_matrix）

举的这个例子，标签label为5类，是个多分类问题

accuracy_score
confusion_matrix

from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score
from sklearn.metrics import precision_score, recall_score, f1_scoreactual = data3['label']     #真实的类别标签（data3是个dataframe，label是其中的一列）
predicted = data3['Label']  #预测的类别标签# 计算总的精度
acc = accuracy_score(actual, predicted)
print(acc)# 计算混淆矩阵
confusion_matrix(actual, predicted)'''
输出结果：0.8969907407407407
array([[1032,    1,    0,    0,    0],[   3,  965,   17,    0,   15],[   0,   10,   42,    0,    6],[   0,   11,    3,    0,    0],[   0,   98,  103,    0,  286]])
'''

2. 使用sklearn计算精确率precision、召回率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

参数average有5个选项：{‘micro’微平均, ‘macro’宏平均, ‘samples’, ‘weighted’, ‘binary’}，默认是default=’binary’二分类

p = precision_score(actual, predicted, average='micro')
p2 = precision_score(actual, predicted, average='macro')
p3 = precision_score(actual, predicted, average='weighted')
#p4 = precision_score(actual, predicted, average='samples')r = recall_score(actual, predicted, average='micro')
r2 = recall_score(actual, predicted, average='macro')
r3 = recall_score(actual, predicted, average='weighted')f1score = f1_score(actual, predicted, average='micro')
f1score2 = f1_score(actual, predicted, average='macro')
f1score3 = f1_score(actual, predicted, average='weighted')
#f1score4 = f1_score(actual, predicted, average='samples')print(p,p2,p3)
print(r,r2,r3)
print(f1score,f1score2,f1score3)'''
输出结果：
0.8969907407407407 0.614528783337003 0.9212413779093402
0.8969907407407407 0.6550877741326565 0.8969907407407407
0.8969907407407407 0.6041619746930149 0.8986671343868164
'''