This feature correlated (cor = 0.4) with our target, however, lasso return no coefficient for any variable.
Our code is like (sklearn)
lassocv2 = LassoCV(cv =features.shape[0], random_state = seed)
lassocv2.fit(features, target)
My data has pasted below.
Dear Wang,
As pointed out by Mohammed, it could be that this input feature is highly correlated with other input features (multicollinearity), and then Lasso just picks one of them. Try Ridge instead of Lasso, and I also leave this old thread which is related to your problem:
https://www.researchgate.net/post/When_creating_a_multiple_regression_equation_what_do_I_do_when_two_of_my_independent_variables_that_I_enter_are_highly_correlated_with_each_other
Best wishes,
Daniel
When covariates are correlated, LASSO typically picks any one of them and ignore the rest. Ridge regression, on the other hand, combines the coefficients of correlated covariates but doesn’t provide sparse models. You can consider the elastic net as it is a compromise between the two.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression