知识卡片

普通读书笔记卡 · 6485.j · 生成模型:GPT-6

拟合误差不能替代未来回测

理解等级:高级

训练历史上的拟合误差描述模型如何解释已有数据,未来预测能力需用按时间留出的区间或滚动回测检验。汇总误差还可能掩盖长尾物品的大误差;本章只给拟合统计,不能据此确认未来预测已达到业务可用标准。

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

      The fitness of the model is provided in terms of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) metrics, as shown in Table 11, which are collected over each Dress˙ID. It is observed that RMSE values are not satisfactory, although the median is likely to be acceptable. How- ever, MAPE values are good, although outliers exist with a higher percentage.

    参考来源

    - 位置:《Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications》第12章《时尚零售评估与预测》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications/_epub-src/OEBPS/Text/9789815136746-C12.html) - 结论依据:摘录中关于拟合误差不能替代未来回测的定义、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:The fitness of the model is provided in terms of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) metrics, as shown in Table 11, which are collected over each Dress˙ID. It is observed that RMSE values are not satisfactory, although the median is likely to be acceptable. How- ever, MAPE values are good, although outliers exist with a higher percentage.