Impact of product structure on stochastic vol

[gver]

I have a data set of autocall trades which has variables like maturity, barrier levels, etc .. when I run a random forest analysis with these categorical variables as independent ones and the stochastic vol premium (stochastic vol minus local vol) as dependent variable then the factor importance chart shows maturity, barrier level , up and out level as the most important feature explaining this variation ..can someone help to understand this intuitively why