Explaining Predictions: Random Forest Post-hoc Analysis (randomForestExplainer package)

We can further evaluate the variable interactions by plotting the probability of a prediction against the variables making up the interaction. The interaction of these two variables are the most frequent interaction as seen in plot_min_depth_interactions. We plot the forest prediction against interactive variables with plot_predict_interaction. However, there is an error when the input supplied is a model created with parsnip. There is no error when the model is created directly from the randomForest package.