Confusion Matrix for 100+ categories

In completing lessons 1 and 2, I wrote a classifier for glyphs in a constructed language called Toki Pona (Toki Pona sitelen pona | Kaggle).

The crux of my problem is that, while I can easily generate the confusion matrix for the 120 categories I’m classifying, it renders so densely that it’s not usable. Has anyone found a solution or alternative to this? I think the visualization would be really useful to have and totally viable for 100s of categories.

1 Like

The display size is adjustable with figsize.


interp.plot_confusion_matrix(figsize=(12, 12))

Thank you! I think I was looking at these docs which didn’t include that arg.