I haven’t really dived into this one yet, but if you’re interested in model interpretability, this might be a good start → https://towardsdatascience.com/introducing-pdpbox-2aa820afd312