search

rikhuijzer / SIRUS.jl

Interpretable Machine Learning via Rule Extraction
https://sirus.jl.huijzer.xyz/
MIT License
31 stars 2 forks source link

Models do not perform well for full range of outcomes #42

Open rikhuijzer opened 1 year ago

rikhuijzer commented 1 year ago

It looks like there is a sort of balancing issue. The following is an image on SIRUS 1.1.2, but the same occurs on 1.3.0:

auc plots

It looks like the model optimizes for one area and not for another. I still suspect this is an issue with the choice of weights. Something is causing the weights to be not chosen properly which then causes sub-optimal classification predictions and poor regression predictions.

rikhuijzer commented 10 months ago

To solve this, I should probably set lambda because the weight logic is very sensitive to the choice of lambda.