Multicollinearity : Should this always be avoided?

supervised_learning
machine_learning

#1

Should I always get rid of multicollinearity? I know this is a assumption for linear regression. Want to understand if I should ignore it for other types of supervised learning methods as well?

Somebody explained multicollinearity to be like trying to judge 2 good lead guitarist playing the instrument at the same time with equally high volume. By this analogy, I think that I should get rid of it everytime irrespective of the modelling technique being used.


#2

Hi @nitish_dydx, please read this post