I am trying to explain my MARS model (for classification). I created a partial dependency plot based on my model: pdp - monthlyIncome

As you can see I have monthly income feature on the x-axis and customer churn probability on the y-axis. Before I train my model, I centered and scaled monthlyIncome feature on the preProcess step. But now in this plot, it doesn't make any sense. so my question is:

  • How can I create this plot with real monthly income values instead of the centered and scaled version?

I am using R. On preprocessing step I used "recipe" package:

blueprint <- recipe(churn ~ ., data = churn_train) %>%
  step_nzv(all_nominal())  %>%
  step_corr(all_numeric(), -all_outcomes(), threshold = .8) %>%
  step_BoxCox(all_numeric(), -all_outcomes()) %>%
  step_scale(all_numeric(), -all_outcomes()) %>%
  step_dummy(all_nominal(), -all_outcomes(), one_hot = TRUE)

and I created partial dependency plot by using "pdp" package:

partial(best_mars_model, pred.var = "monthlyIncome", prob = TRUE) %>% autoplot()

Thanks for help


Your Answer

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

Browse other questions tagged or ask your own question.