Add Adaptive Splitting Decision Trees to River #1609
danielnowakassis
started this conversation in
Show and tell
Replies: 2 comments
-
Summoning the resident tree hugger @smastelini 🌴🤗 |
Beta Was this translation helpful? Give feedback.
0 replies
-
Hi @danielnowakassis , sounds good to me! The idea also sounds really clever :) |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi.
I've implemented my decision tree in River and run all the tests for adding a new estimator.
The published paper is here: https://dl.acm.org/doi/10.1145/3605098.3635899
We proposed an adaptive splitting mechanism, where a change detector monitors the error rate or data distribution purity (parameter) of the leaf node to determine the split point (and if merit > 0). It had some good results against SOTA decision trees, and I have also tested it in real datasets in River.
Accuracy :
I am now writing an ablation study of the method. I'm open to any questions. LAST would be error rate monitoring, LAST_D would be data distribution purity monitoring.
Can I do a PR?
Beta Was this translation helpful? Give feedback.
All reactions