Multi collinearity and unnecessary features removing using regularization


I have a doubt wrt the extent of application of regularization that it handles the data which has so much of junk(unnecessary features-- like a phone number feature to predict house price) or it handles features which is highly related(multi collinearitty) or high degree polynomials like x,x^2,x^3 and so on… or both??

I am a bit confused on the above situation as we have other methods to remove the un necessary features like pearson coffieceint,LDA,chi-square. can the regularization replace those methods in any scenario?

Hi @annem,

there are various dimensionality reduction techniques like you mentioned. And are used in different scenarios. I would like to point out that chi square and pearson coeficient are not exaclty for removing features. for example, chi square is to find out the relationship between categorical variables, or how correlated they are.

© Copyright 2013-2019 Analytics Vidhya