How to tell a my prediction model no to evaluate a feature if the value of the feature is 0

Question

I am using a catboost regressor to predict the impact of several independent variables on a dependent variable. My dependent variable is binary, my independent variables are not