Fazer estudo para adaptação de PU Learning

[ryukinix]

Estudo e planejamento estrutural de como aplicar PU Learning para o nosso problema.

Survey: https://link.springer.com/article/10.1007/s10994-020-05877-5

lib: https://pulearn.github.io/pulearn/

[Oscar280578]

I found this article with a simple python code. He uses an xgboost model, but I think that catboost may be used too. https://medium.com/@aniruddha.am.mitra/decoding-pu-learning-a-fusion-of-statistics-and-machine-learning-8709c35d0352 Note that for implementing the PU learning, we have to do a copy of dataset. I think that we can adapt this code to run with a catboost model.

[Oscar280578]

In this site, we can observe an example of Elkan and Noto (a type of PU learning) used for cancer dataset, like the problem that we are tackling. https://github.com/pulearn/pulearn In this case, he used the randon forest, but i think that the code may be adapted to incorpore a catboost model.

[ryukinix]

Very interesting finds, Oscar! I'll take a time as soon as possible to read what you pointed here.

[ryukinix]

Considerando nossos alinhamentos, vamos considerar essa tarefa feita, como um estudo superficial sobre o conceito com algumas referências. @helen0l @Oscar280578