Difference between balanced_accuracy_score and accuracy_score

Question

I am using balanced_accuracy_score and accuracy_score both in sklearn.metrics.

According to documentation, those two metrics are the same but in my code, the first is giving me 96% and the second one is 97% while accuracy from training is 98%

Can you explain to me what is the difference between the three accuracies and how each is computed?

Note: the problem is a multi-classification problem with three classes.

I have attached code samples.

accuracy is 98%

model.compile(loss='categorical_crossentropy',
              optimizer=Adam(lr=0.00001),
              metrics=['accuracy'])

accuracy is 96%

from sklearn.metrics import balanced_accuracy_score
balanced_accuracy_score(all_labels, all_predications)

准确率为 97%

from sklearn.metrics import accuracy_score
accuracy_score(all_labels, all_predications)

Answer 1

据我了解的问题（不知道什么all_labels，all_predictions）在你OUT之间样本外预测的运行上的差异balanced_accuracy_score以及accuracy_score由前者功能的平衡造成的。

accuracy_score 简单地返回您正确预测的标签百分比（即有 1000 个标签，您准确预测了 980 个，即您获得了 98% 的分数。

balanced_accuracy_score但是工作方式不同，因为它返回每个类的平均准确率，这是一个不同的指标。假设您的 1000 个标签来自 2 个类，其中第 1 类中有 750 个观察值，第 2 类中有 250 个观察值。如果您在每个类中预测错误 10 个，则第 1 类的准确率为 740/750= 98.7%，而 240/250=96第 2 类中的 %balanced_accuracy_score将返回 (98.7%+96%)/2 = 97.35%。因此，我相信根据文档，该程序可以按预期工作。

Answer 2

Accuracy = tp+tn/(tp+tn+fp+fn) 不适用于不平衡的类。

因此我们可以使用平衡精度 = TPR+TNR/2

TPR= 真阳性率 = tp/(tp+fn) ：也称为“敏感度”

TNR = 真阴性率 = tn/(tn+fp) ：也称为“特异性”

Balanced Accuracy 给出的结果几乎与 ROC AUC Score 相同。

链接：

1 https://en.wikipedia.org/wiki/Precision_and_recall

2 https://scikit-learn.org/stable/modules/generated/sklearn.metrics.balanced_accuracy_score.html#sklearn.metrics.balanced_accuracy_score

3 https://scikit-learn.org/stable/modules/generated/sklearn.metrics.roc_auc_score.html