知识卡片
精确率召回与 AUC
精确率与召回率都围绕“正例”(如垃圾邮件):精确率问“判为正例时有多少是真的”,召回率问“全部正例里找出了多少”。类别极不均衡时它们比准确率稳。两者存在张力:提高阈值则精确率升、召回率降,降低阈值则相反,必须一起看。AUC ROC 是阈值无关的:画各阈值下真正率与假正率的曲线求面积,等价于“随机取一正一负,模型给正例更高分的概率”;但它会被远离合理阈值的排序改进欺骗。类别高度不平衡时 PR 曲线下面积更有信息,尽管没有严格理论含义。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- Evaluating Model Validity and Quality · 查看原书依据
Conceptually, AUC ROC is computed by creating a plot showing the true-positive rate and the false-positive rate for a given model at every possible classification threshold, and then finding the area under that plotted curved line; see Figure 5-3 for an example.
参考来源
- 位置:《Reliable Machine Learning》第05章《Evaluating Model Validity and Quality》(源文件:extracted/Reliable Machine Learning/_epub-src/OEBPS/ch05.xhtml)
- 结论依据:原文明确阐述该主题,并给出定义、实践或反例。
- 原始内容:Conceptually, AUC ROC is computed by creating a plot showing the true-positive rate and the false-positive rate for a given model at every possible classification threshold, and then finding the area under that plotted curved line; see Figure 5-3 for an example.