scikit准确率召回率F1模块

from keras.models import load_model
import numpy as np
from sklearn.metrics import confusion_matrix, f1_score, precision_score, recall_score,accuracy_score
 
 
# 测试
# model=load_model('./models/lenet5_weight.h5')
 
y_predict = model.predict(test_images, batch_size=512, verbose=1)
# y_predict = (y_predict > 0.007).astype(int)
y_predict = (y_predict > 0.01).astype(int)
y_true = np.reshape(test_labels, [-1])
y_pred = np.reshape(y_predict, [-1])
 
# 评价指标
accuracy = accuracy_score(y_true, y_pred)
precision = precision_score(y_true, y_pred)
recall = recall_score(y_true, y_pred, average='binary')
f1score = f1_score(y_true, y_pred, average='binary')
 
# Micro F1: 将n分类的评价拆成n个二分类的评价，将n个二分类评价的TP、FP、RN对应相加，计算评价准确率和召回率，由这2个准确率和召回率计算的F1 score即为Micro F1。
# Macro F1: 将n分类的评价拆成n个二分类的评价，计算每个二分类的F1 score，n个F1 score的平均值即为Macro F1。
# 一般来讲，Macro F1、Micro F1高的分类效果好。Macro F1受样本数量少的类别影响大。
micro_f1 = f1_score(y_true, y_pred,average='micro')
macro_f1 = f1_score(y_true, y_pred,average='macro')
 
 
print('accuracy:',accuracy)
print('precision:',precision)
print('recall:',recall)
print('f1score:',f1score)
print('Macro-F1: {}'.format(macro_f1))
print('Micro-F1: {}'.format(micro_f1))