How can a reusable ColumnTransformer select all numeric columns without hard-coding their names?
Classic MLLevel: IntermediateDtype-based column selection
How can a reusable ColumnTransformer select all numeric columns without hard-coding their names?
Hints
- Selection can be a callable.
- Storage dtype is not semantic type.
Answer
Use a column selector based on dtype, such as make_column_selector(dtype_include=np.number). Still validate the input schema: an identifier stored as an integer is not necessarily a useful numeric measurement.
Source and licence: scikit-learn 1.7.2 — compose · BSD-3-Clause · answer: Quantoros editorial