I have studied how to build PDP plot for machine learning models. For traditional stat model, normally we just do a simple bi-var plot between x and y(e.g. rank x and cut it into 10bins, for each bin, calculate mean x, mean y, then plot) , and PDP looks just like a bi-var x-y plot.
Why for ML models, it has to replicate datasets for multiple times to do PDP plot ?
Thanks