SGDRegressor worse than OLS?

Machine Learning with Python: Zero to GBMs Lesson 1 - Linear Regression with Scikit Learn
#1

Hi everyone,
I just did the exercise with SGDRegressor, the results are a little worse than the simple OLS. Is that normal? And if yes, why is that?

Thanks a lot!

#2

That means your model is not good and you need to do more hyperparameter-tuning/feature selection for your model. Maybe some columns you are selecting does not have a linear relation with the target but you are choosing a linear model. You, can try using different approaches to solve the problem.
Yes it is completely normal. Cause this is the most basic thing you have done without feature selection/data processing, hyperparameter tuning, and used the most basic model which might not be the best model you have chosen.

2 Likes
#3

Thanks birajde for the help. Really enjoy Jovian and this learning experience!

1 Like