0
$\begingroup$

I've bee using R for predictive analytics and here is the issue: I'm trying to predict the species (E1, E2, E3 and E4) of an animal using as predictors a set of categorical (factors) (NO1, NO2, NO3, NO4 and NO5) and numeric (NU1, NU2 and NU3) variables. For that I used a few different methods: Naive Bayes, Decision Trees, Support Vector Machines and Multinomial Logit.

I'd like to know if there is a way to evaluate (quantify or qualify) the importance/impact of each variable/predictor I use for all the methods I mention above or at least for one of them, for instance that, say, NO3, NO4 and NU2 aren't good in predicting the species but the other predictors are.

Many thanks in advance.

Diego.

$\endgroup$

migrated from stackoverflow.com Jun 2 '15 at 4:18

This question came from our site for professional and enthusiast programmers.

  • $\begingroup$ Please read the help pages for what sorts of questions are on-topic ar SO. $\endgroup$ – DWin Jun 2 '15 at 3:54
0
$\begingroup$

Indeed! you can use forward selection, backward selection, dimensionality reduction, penalization etc. Take a look at chapter 6 here:

http://www.dataschool.io/15-hours-of-expert-machine-learning-videos/

Don't miss the lab: best subset selection which is also in R! -Best of Luck!

PS the free pdf of the course can be found here: http://www-bcf.usc.edu/~gareth/ISL/

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.