how the training/fitting happens when both x_train and y_train are of different sizes?

[teddcp2]

hi all,

I have a doubt. It's like I have divided the data by calling train_test_split of SKLEARN. Then I am removing the outliers using percentile method for training set only.

So now my x_train is reduced compared to what it had.

But here I have question i.e. even if my y_train still has the same old size and x_train has a different size, My model is getting trained. Won't it be wrong to let the model fitting the data ?

If not, how the model is accompanied in the change?

[solegalli]

Yes, if you remove outliers from the train set using the Outlier trimmer, then you need to remove from y_train those observations as well.
You can do so my looking at the X_train and y_train indeces.

[teddcp2]

Hi @solegalli , Thanks for answering. Yep, I understood your point. Thanks a lot.

Just one last question. I was going through your cross-validation example and let's say i want to do SKLEARN CV with OutlierTrimmer.

It won't work right? Because of the size mismatch between training and test set.

To do a CV with OutlierTrimmer, manually need to write a code to iterate over the data n-fold times and split data , remove outliers and handle missing values for training set, finally measure the score for test set.

Let me know if my above approach is correct or not. Thanks in advance :)

But with Winsorizer, SKLEARN CV will work.right ? As there wont be any size mismatch issues.

[solegalli]

that is correct