Relevant attributtes in lasso

Anyone know how I can find the relevant attributes of a data set from the model coefficients lasso? That is, I have to pay attention, the largest coefficients? the largest absolute value?

asked May 03 '13 at 12:33

byfnk
21●3●3●6

edited May 03 '13 at 12:43

As I know, when variables have high correlation, LASSO will show randomness, that is, it will select one of the correlated variables randomly.

(May 03 '13 at 14:34) Ziyuan Lin

Thanks for your answer! It is a question for a faculty homework machine learning. I fit the linear and lasso and pick up the coefficients associated and ask me to determine which are the most characteristic attributes (so the most important)

(May 03 '13 at 22:59) byfnk

Is it asking you to pick up coefficients relevant to the response?

(May 04 '13 at 12:02) Ziyuan Lin

One Answer:

oldest newest most voted

The idea of lasso is to bring non-relevant attributes' coefficients to zero, so you may be interested in non-zero coefficients. More here:

http://fastml.com/large-scale-l1-feature-selection-with-vowpal-wabbit/

answered May 05 '13 at 18:28

Zygmunt Zając
46●1●1●4

thanks this was the answer.

(Jun 12 '13 at 12:35) byfnk

Subscription:

Once you sign in you will be able to subscribe for any updates here

Tags:

regression ×101
scikit-learn ×21
lasso ×13

Asked: May 03 '13 at 12:33

Seen: 667 times

Last updated: Jun 12 '13 at 12:35

Relevant attributtes in lasso

Subscription:

Related questions