Variables that are important in random forest don't necessarily have any sort of relationship with the outcome. So would it be wise to use random forest to gather the most important features and then plug those features into a poisson regression model?If not, what other approach can be used for feature selection?(independent variables are categorical and dependent variable is count).

  • 1
    $\begingroup$ How many independent/feature variables and how many cases do you have? Those details would help provide a more useful answer. $\endgroup$
    – EdM
    Oct 31, 2018 at 17:43
  • $\begingroup$ I have 61 features out of which 51 are categorical variable and 10 are count variables. $\endgroup$
    – Trupti J
    Oct 31, 2018 at 18:21