I know that I can link several estimators that implement transformation methods to transform X (feature set) in sklearn.pipeline. But I have a use case where I want to transform the target label (such as Convert the label to [1… K] instead of [0,K-1], I am happy to use it as a component in my pipeline. Is it possible to use sklearn.pipeline.?
No, the pipeline will always go through the same. Transformation is performed outside the pipeline.
(This is the scikit-learn A known design flaw, but it has never had enough pressure to change or extend the API.)
I know that you can link several estimators that implement transformation methods to convert X (feature set) in sklearn.pipeline. But I have a use case where I want to transform the target label (for example, transform the label to [1… K] instead of [0, K-1], which I am happy to use as Is it possible for a component in my pipeline to use sklearn.pipeline.?
No, the pipeline will always go through unchanged. Transformation is performed outside the pipeline.
< /p>
(This is a known design flaw in scikit-learn, but it never has enough pressure to change or extend the API.)