Maybe there is a free tool or library that can help here. I have a working model that predicts customer churn. Now like any person would ask, what do I do to reduce that probability of churn. My mind thinks grab the top 3 feature importance from the model and flip the value 1->0 or 0->1 then send the new data frame back into the model for each customer, then do the same for the other two importance's. Then take the best value and output that as change suggestion.
I don't know, I feel like there has to be something better. Any suggestions welcome as I'm relatively new in the python world.
