OneHotEncoder() drop='first'

[shashwat9kumar]

Hey, I read at many online sites that using drop='first' will reduce redundancy and help the model learn faster and better.
But when I implemented it, the model gave a higher MAE and MSE.
Is there something I might be doing wrong?

[ashikshafi08]

Hey, @shashwat9kumar It might be true for instances where your problem suffers from multicollinearity, where more than 1 independent variable is related to your dependent variable.
In this case drop = "first" will work during encoding it reduces the problem of multicollinearity by a minimum level.

If your problem is multicollinearity? then you can use drop = first if not the default would work. This is the one reason I know and there might be n number of reasons to use drop = 'first'.

Tbh the documentation says the same, I attached a pic below

Hope it helps!

[shashwat9kumar]

Hey @ashikshafi08
Thanks so much!!! It's a really great explanation