知识卡片
缩放不能代替清洗或安全控制
Min–Max和标准分数改变数值尺度,不自动去重、补缺失、修数据错误或增强保密。新数据超出训练范围时Min–Max可产生区间外值,是否截断还需另作有据选择。
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
- · 使用数据
原书出处 · 1 处
- · 查看原书依据
Data normalization reorganizes the data so that it can be used uniformly and is similar for further analysis. It is a process of data cleaning by eliminating redundant and unstructured data and making the data appear uniform across all features. To normalize a data set, the data need to be scaled to a common range, such as between 0 and 1. This is done using
参考来源
- 位置:《Machine Learning Fundamentals: Concepts, Models, and Applications》第05章《使用数据》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Machine Learning Fundamentals: Concepts, Models, and Applications/_epub-src/OEBPS/xhtml/17_chapter05.xhtml)
- 结论依据:摘录中关于缩放不能代替清洗或安全控制的定义、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:Data normalization reorganizes the data so that it can be used uniformly and is similar for further analysis. It is a process of data cleaning by eliminating redundant and unstructured data and making the data appear uniform across all features. To normalize a data set, the data need to be scaled to a common range, such as between 0 and 1. This is done using
- 生成时纠错核验:[官方文档](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html)(2026-11-06;定义与接口职责,未执行书中示例)。