added _classes for knn classifiers #679
Conversation
supervised/algorithms/knn.py
Outdated
|
|
||
| class KNeighborsAlgorithm(KNNFit, RegressorMixin): | ||
| class KNeighborsAlgorithm(KNNFit, ClassifierMixin): |
There was a problem hiding this comment.
this looks strange, we derive from ClassifierMixin in KNNFit as well ...
supervised/algorithms/knn.py
Outdated
| @@ -73,7 +81,7 @@ def __init__(self, params): | |||
| ) | |||
|
|
|||
|
|
|||
| class KNeighborsRegressorAlgorithm(KNNFit): | |||
| class KNeighborsRegressorAlgorithm(KNNFit, RegressorMixin): | |||
There was a problem hiding this comment.
we derive from KNNFit which derives from ClassifierMixin and we also derive from RegressorMixin. This looks strange ...
There was a problem hiding this comment.
i derived the KNeighborsAlgorithm from both ClassifierMixin and KNNFit, where as the KNeighborsRegressorAlgorithm from both RegressorMixin and KNNFit .
KNNFit focuses on fitting the model, and KNeighborsAlgorithm and KNeighborsRegressorAlgorithm focus on the specific tasks of classification and regression, respectively.
tests/tests_algorithms/test_knn.py
Outdated
| classes = None | ||
|
|
||
| #debug statements | ||
| print("np.unique(self.y):", np.unique(self.y)) |
There was a problem hiding this comment.
we dont need debug statements, let use asserts
|
made the requested changes, and its working quite great.
|
|
@pplonski is this working? |
|
Thank you @Hk669 for PR, good job 🥇 |
|
Thanks for the support @pplonski |
Added _classes attribute for k-Nearest Neighbors (KNN) Algorithm
Description:
This pull request addresses the issue #669 and implements the addition of the _classes attribute for the k-Nearest Neighbors algorithm.
Changes Made:
Modified the KNNFit class to include the _classes attribute, providing unique classes based on the fitted model.
Updated the KNeighborsAlgorithm and KNeighborsRegressorAlgorithm classes to inherit from KNNFit and ClassifierMixin and RegressorMixin, respectively.
Adjusted relevant unit tests to ensure the proper functioning of the _classes attribute.
Testing:
checklist: