Two classification indicators Precision, Recall, Accuracy, F score, TPR, FPR, MCC

1. Ture Positive Rate (TPR)：$\frac{TP}{TP+FN}$, to examine the proportion of correct predictions of the classifier in GT-positive samples.
2. False Positive Rate (FPR)：$\frac{FP}{FP+TN}$, to examine the proportion of classifiers predicting errors in GT-negative samples.
3. AUC: The area under the ROC curve, which comprehensively examines the performance of the classifier.
4. Percision：$\frac{TP}{TP+FP}$, to examine the proportion of correct predictions among the samples predicted by the classifier as positive.
5. Recall：$\frac{TP}{TP+FN}$, same as TPR.
6. Average Precision (AP): The area under the PR curve, which comprehensively examines the performance of the classifier.
7. ACC：$\frac{TP+TN}{TP+FP+FN+TN}$, to examine the proportion of correct predictions of the classifier in all samples.
8. F1 score：$\frac{2}{\frac{1}{\text{ Precision }}+\frac{1}{\text{ Recall }}}$, to calculate the harmonic mean of the two indicators of Precision and Recall, taking into account the performance of the classifier on these two indicators.
9. F $\beta$ score：$\frac{1+b^2}{b^2\frac{1}{\text{ Recall }}+\frac{1}{\text{ Precision }}}$, compared to F1 score, F $The\; \beta$ score can adjust the weight to make the indicator pay more attention to the performance of the classifier on Recall or Precision (for example, F2 score pays more attention to Recall).
10. Matthews correlation coefficient (MCC)：$\frac{TP\times TN-FP\times FN}{\sqrt{(TP+FP)(TP+FN)(TN+FP)(TN+FN)}}$​, considering the four situations of TP, FP, FN, and TN comprehensively, this indicator can be used even when the sample content of the two categories is very different. (I didn't understand this indicator very well, so I found the meaning of this indicator from the Internet)