Hands-On Data Analysis with Pandas

Creating the PartialFitPipeline subclass

We saw in Chapter 9, Getting Started with Machine Learning in Python, that the Pipeline class made streamlining our machine learning processes a cinch, but unfortunately, we can't use it with the partial_fit() method. To get around this, we can create our own PartialFitPipeline class, which is a subclass of the Pipeline class, but supports calling partial_fit(). The PartialFitPipeline class is in the ml_utils.partial_fit_pipeline module.

We simply inherit from sklearn.pipeline.Pipeline and define a single new method—partial_fit()—that will call fit_transform() on all the steps except the last one, and partial_fit() on the last step:

from sklearn.pipeline import Pipelineclass PartialFitPipeline(Pipeline): ...