You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am plotting XGBClassifier trees using bare dtreeviz.model().view() - below is a screenshot of one of the trees.
The first leaf has 10 to 100 times lower entries, but it is impossible to see the difference from the radius of the piechart.
Looking at the examples, it seems to be implemented, but I was wondering whether this is a bug or just a scaling issue. The dataset sizes in the examples are much smaller than mine. I think it is just a scaling issue - after a certain threshold(around 100), the radius doesn't increase, as can be seen in a different tree below.
One could implement logarithmic scaling instead of linear scaling of the pie chart radius to improve the situation.
I will definitely look at the source and try to figure it out; I just wanted to share it here. I was also wondering if there are better visualizations of the leaves for large datasets.
Thanks again,
Sayan
The text was updated successfully, but these errors were encountered:
Hello.
Thank you again for making this repo public!
I am plotting XGBClassifier trees using bare
dtreeviz.model().view()
- below is a screenshot of one of the trees.The first leaf has 10 to 100 times lower entries, but it is impossible to see the difference from the radius of the piechart.
Looking at the examples, it seems to be implemented, but I was wondering whether this is a bug or just a scaling issue. The dataset sizes in the examples are much smaller than mine. I think it is just a scaling issue - after a certain threshold(around 100), the radius doesn't increase, as can be seen in a different tree below.
One could implement logarithmic scaling instead of linear scaling of the pie chart radius to improve the situation.
I will definitely look at the source and try to figure it out; I just wanted to share it here. I was also wondering if there are better visualizations of the leaves for large datasets.
Thanks again,
Sayan
The text was updated successfully, but these errors were encountered: