← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9386.b · 生成模型:GPT-6

综合分数不能代替业务错误代价

理解等级:中级

F1为2PR/(P+R),本例为2/3;准确率是(TP+TN)/总数,为0.6。F1不含真负类,且未必反映全部误判成本,选择指标和阈值应根据任务,不是默认越接近1越足以部署。

这张卡片看不懂?

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

    • · 查看原书依据

      A common metric used by data scientists that takes both precision and recall into account is F1 score. As the following listing shows, it’s calculated using the harmonic mean of precision and recall—their product divided by their sum.

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》附录D《附录D:机器学习工具与技术》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/appendix-d.html) - 结论依据:摘录中关于综合分数不能代替业务错误代价的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:A common metric used by data scientists that takes both precision and recall into account is F1 score. As the following listing shows, it’s calculated using the harmonic mean of precision and recall—their product divided by their sum.