https://tracholar.github.io/machine-learning/2018/01/26/auc.html
https://blog.csdn.net/qq_22238533/article/details/78666436
y轴: tpr = tp/p
x轴: fpr = fp/n
正样本中取出样本预估为正的概率 > 负样本中取出负样本为正的概率
---->
正样本排在负样本前面
---->
排序性好。
![](https://img.haomeiwen.com/i1560080/2e1218f3f3d3a808.png)
image.png
计算auc
https://blog.csdn.net/qq_22238533/article/details/78666436
![](https://img.haomeiwen.com/i1560080/b6f8f7944b4395df.png)
![](https://img.haomeiwen.com/i1560080/8817652255d69ca1.png)
![](https://img.haomeiwen.com/i1560080/39bd85d22f704e44.png)
roc 曲线
纵轴刻度间隔设为1/p,横轴设为1/n;按照score倒排,遇到真实正例y+1,遇到真实负例x+1
https://stats.stackexchange.com/questions/145566/how-to-calculate-area-under-the-curve-auc-or-the-c-statistic-by-hand
![](https://img.haomeiwen.com/i1560080/e8bb9f9fd08c51a8.png)
网友评论