If you label-encode something which is qualitative, like brand of toothpaste or colour of hair, would you describe the resulting data as quantitative since it is now expressed in numbers? Or would you continue to describe the data as qualitative, since it is still describing the same underlying information?
Mapping to numbers in a random fashion e.g. Label encoder doesn't make it quantitative.
It's still qualitative.
It just that most of the library implementation of ML Models only expect numeric data.
So, Label encoding will work fine with models that don't do any quantitative operations on any two encoded values i.e. Tree-based models that simply use the label to divide the space
If you use it with models that treat the mapping as quantitative e.g. Neural Networks then the result might be error-prone e.g. 10 will become 2 times 5 although no such relationship was intended.
Answered by 10xAI on December 6, 2020
Get help from others!