search

scikit-learn-contrib / DESlib

A Python library for dynamic classifier and ensemble selection
BSD 3-Clause "New" or "Revised" License
479 stars 106 forks source link

Pool based on different features #218

Closed jayahm closed 3 years ago

jayahm commented 4 years ago

Hi,

Is there any example on how to perform DS based on different features?

The examples on the website are only cover on pool of heterogeneous classifiers, bagging and random forest.

I couldn't find and have no idea on how to generatea pool of different features.

Menelau commented 4 years ago

@jayahm Hello,

At the moment the library can only work with all models receiving the same feature set as input. There are ways to have base classifiers using different transformations of the input features set by using a sklearn pipeline (for example use a subset of the input features or transformations such as PCA). In this case, each base model should be composed of a scikit-learn Pipeline in which the first component is the transformation applied to the input features and the last component is the classifier model.

I will create an example showing this functionality.

jayahm commented 4 years ago

How about Random Subspace Method?

I was thinking to try this.

Menelau commented 3 years ago

fixed in #254

jayahm commented 2 years ago

Hi,

I just realized the issue has been fixed. But, I'm not sure how to start applying it