Your model doesn't appear to be balanced?

Question

When executing:
y_test_pred = clf.predict(x_test_transf)
I get this error:

ValueError: X has 1646 features, but MultinomialNB is expecting 3810 features as input.

Which appears to mean that the Multinominal is expecting a shape for the model that is different from what the shape of the model is. In other words, it's not balanced.
As much as I would like to try to balance the model myself. I think I would do more harm than good so if you could either balance the model or tell me where I'm going wrong it would be most appreciated.
Thanks..

Answer 1

Hey Timothy,

Thank you for your question!

This error might be caused by something that you have done up in the code. Unfortunately, I cannot reproduce the error you are getting and cannot help you with a solution. What I can suggest is to restart the notebook from Kernel -> Restart and run everything anew. Note that it is important to execute the cells in order.

Kind regards,
365 Hristina

Answer 2

So I did get your notebook to work finally. I'm going to have to go over what I did. Thank you so much for the confirmation that I was doing something wrong.

Answer 3

Hello, I think the problem is that you used vectoriser.fit_transform(x_test) on the x_test variable, instead of just vectoriser.transform(). The x_test variable has imbalanced features because you fit the data on the vectoriser, and then you transform it. Using only transform makes it so that the x_test variable is transformed by the vectoriser that is fit on the x_train variable

Answer 4

I am getting the error below, not sure what i am doing wrong

AttributeError                            Traceback (most recent call last)
Cell In[93], line 1
----> 1 ConfusionMatrixDisplay.from_predictions(
      2 y_test, y_test_pred,
      3 labels = clf.classes_,
      4 cmap = 'magma');

AttributeError: type object 'ConfusionMatrixDisplay' has no attribute 'from_predictions'

Answer 5

Hey Erick,

Thank you for reaching out!

Please, update your scikit-learn library to version 1.0 or higher. This will solve the issue.

Kind regards,
365 Hristina

Your model doesn't appear to be balanced?

Submit an answer