Please let me know if I can use any visualisation technique for XGboost similar to that of Multidimensional scaling plot of proximity from random forest. thanks!
Asked
Active
Viewed 16 times
1
-
This is my first time reading about MDS plots (see https://stats.stackexchange.com/q/40737/232706), but mechanically it should work in exactly the same way: the proximity matrix still works for any tree ensemble, and then you take principal components. For a boosted ensemble though, it's not clear that the impact on proximity should be uniform across all the trees; being separated by early trees seems to be "more important" than being separated by later trees? – Ben Reiniger Oct 29 '21 at 14:33