How can we apply model oriented ibreakdown to complete data set

[anuradham7]

right now it just takes one observation at a time, is there any provision to run it for entire data set ? bd_rf <- predict_parts(explainer = explain_rf,new_observation = test[1,] , type = "break_down")

[hbaniecki]

Hi, did you consider looping over the observations of interest? This is how I apply it in the modelStudio package.

[pbiecek]

the reason was that the plot() function would not draw explanations well with predict_part() for multiple points. With predict_profile(), you can draw all the profiles in one plot. But for predict_part() there is no good way to simultaneously draw multiple attributions for multiple observations.